Virtual services

Politecnico Di Milano (POLIMI)

Providing expertise in the field of medical robotics (medical imaging for surgical planning, virtual reality, extended reality) to be eventually tailored to specific projects: 1) Definition of applications objectives 2) Data collection, data curation and bias evaluation 3) Preprocessing and optimization of feature extractions 4) Selection of suitable machine learning models, model training and testing 5) External validation 6) Interpretations of models (XAI algorithm applications and development) Keywords: AI, algorithm development, AI testing and validation, Medical Imaging for Surgical Planning, Virtual Reality, Extended Reality

Fondazione Bruno Kessler (FBK)

This service covers all key aspects of development and deployment of AI solutions, based on shallow machine learning or deep learning). State-of-the-art pipelines are generated/improved leveraging on various technonolgies or solutions such as: data encoding, management of missing data, data augmentation and syntetic data generation, model selection and optimisation, hyperparameters tuning, fine-tuning, data shift and transfer learning, reproducibility and explainability. Keywords: Biomedical data Analysis, Deep Learning, CNN, Machine Learning, Image Analysis, Synthetic Data Generation

Multitel (MULTITEL)

Design, Optimization and Small Series Production of Prototypes: The MULTITEL team excels in creating bespoke prototypes and limited series, demonstrating the potential of AI and optoelectronics applications in various domains.

Laboratoire National De Metrologie Et D'Essais (LNE)

Evaluation of the AI system regarding its robustness against cybersecurity issues ( risk assessment, secure Data, access Control and Authentication, etc …) This will include the design of test protocols, the realization of tests, the analysis of results and production of a test reports.

Karolinska Institutet (KI)

## Overview This service supports SMEs and researchers in developing, training, and validating AI models for medical imaging applications. Hosted at SMAILE, Karolinska Institutet, it covers segmentation, detection, and classification tasks using CT, MRI, nuclear images, multimodal images, ultrasound, histology, and microscopic images. The pipeline includes data curation, evaluation of labeling strategy, model selection and training, evaluation metric selection, and model performance benchmarking. We provide expertise and support in: - AI model training and optimization for medical imaging - Validation using clinical datasets and standard metrics (using publicly available datasets, datasets available through data agreements, and internal datasets at KI, depending on the case) - Clinical Relevance and Comparison with the state of the art in research and clinical practice - Imaging biomarkers studies for diagnosis, prognosis, and prediction applications ### How can the service help you? The service ensures your imaging AI model performs reliably and is aligned with clinical expectations. Whether you’re entering the pre-clinical testing phase or seeking validation to secure investment or regulatory approval, this service equips you with a rigorous evaluation and feedback report. ### How the service will be delivered? Available both virtually and physically. Imaging data can be reviewed remotely through a secure data transfer process. On-site collaboration is also possible for sensitive datasets or model development, evaluation, and validation. A typical project takes 4–6 weeks. --- ## Additional information ### Provider description SMAILE is the digital health core facility at Karolinska Institute. It offers interdisciplinary support in AI for medical imaging, data analytics, and system validation, partnering with leading institutions under the Swedish TEF-Health node. ### Technical description The imaging model pipeline is extensively validated by using a curated benchmark set and standard evaluation frameworks such as well-established quantification metrics for object detection, image segmentation, and classification tasks.. Annotation quality is reviewed, and model performance is benchmarked against open or reference models. Standard medical image processing tools such as PyDicom, ANTs, and ITK, as well as community-driven open-sourced frameworks such as MONAI, are the core components of our designed pipelines. ### Service customization The service can focus on either model development, evaluation, or both. Datasets can be anonymized and securely shared, or analysis can be conducted in a local sandboxed environment.

RISE Research Institutes Of Sweden

Method description: Develop and improve AI algorithms utilizing Deep Learning paradigms within Computer Vision, Data Analysis and Natural Language Processing. The service can include suggestions of applications and highlight strengths, weaknesses, or software development. The service is customized according to the SME's needs. Pre-requisites for the service are the analysis of data availability, data-readiness level and customer needs. Method reference: To be refined depending on the type of task, and in dialog with the customer. The Center for Applied AI at RISE carries out cutting-edge research in AI, connects expertise and applications within RISE, and explores a wide range of innovative applications with industry and the public sector. Applied AI Centre at RISE helps companies and government agencies to see more potential in the technology, use it more wisely and develop it faster. Keywords: Trustworthy AI, Federated learning, Data Science, AI, Deep Learning, Natural language processing, Computer Vision, Machine Learning

Multitel (MULTITEL)

Development and improvement of innovative AI solutions, tailored to the unique challenges and opportunities of the SME. From 1D, 2D, 3D data or multimodal data, MULTITEL develops and improves customized AI models, leveraging its expertise in AI, embedded systems, machine learning, optoelectronics and optics to deliver robust, efficient solutions. With a focus on refinement, MULTITEL also enhances the performance of existing AI models, ensuring they meet the latest standards in speed, accuracy, and functionality.

Unidade Local De Saúde De Coimbra EPE (ULS Coimbra EPE)

The service offers SMEs expert evaluation and validation of their AI models intended for clinical use. Leveraging the hospital's domain expertise in healthcare and data analytics, this service assesses the accuracy and clinical suitability of AI models in real-world clinical scenarios by assessing models against clinically relevant metrics, benchmarks, and regulatory standards, it ensures their safety, reliability, and effectiveness in real-world healthcare settings.

Centro Hospitalar De Sao Joao Epe (CHSJ)

The service offers SMEs expert evaluation and validation of their AI models intended for clinical use. Leveraging the hospital's domain expertise in healthcare and data analytics, this service assesses the accuracy and clinical suitability of AI models in real-world clinical scenarios by assessing models against clinically relevant metrics, benchmarks, and regulatory standards, it ensures their safety, reliability, and effectiveness in real-world healthcare settings.

Fraunhofer Gesellschaft Zur Forderung Der Angewandten Forschung Ev (Fraunhofer)

**Who Can Benefit:** - MedTech & Digital Health Companies: For developing high-performance AI models on sensitive clinical data without cloud dependency or data sovereignty concerns. - Research Groups: For leveraging state-of-the-art computing infrastructure and ML expertise to accelerate model development. - Startups: For accessing enterprise-grade training infrastructure and clinical AI expertise without heavy upfront investment. **Key Features:** - End-to-end ML/DL pipeline development: data preprocessing, feature engineering, model architecture design, training, and hyperparameter optimization - Support in developing RAG Systems and Agentic AI: on-premise infrastructure for deployment and limited fine-tuning. - Secure on-premises computing clusters at the highest data security standards – no cloud computing required - Access to the High-Performance Computing (HPC) cluster at FAU Erlangen-Nürnberg for large-scale training workloads - Rigorous testing methodology including cross-validation, holdout testing, and subgroup performance analysis - Clinical domain experts from affiliated clinics ensure models are grounded in medical reality and provide valuable feedback for algorithm improvements **Possible Applications:** - Digital Biomarker Development (e.g. Cardiology): Training models to detect arrhythmias, atrial fibrillation, or heart failure decompensation from wearable ECG or PPG signals. - Biomechanical Movement Classification: Building AI models for automated gait pattern recognition, fall risk scoring, or joint load estimation from inertial sensor data collected in motion labs. - Neurological Disorder Detection: Developing classifiers for tremor subtypes, seizure prediction, or cognitive decline indicators based on neurophysiological signals. - Athletic Performance Optimization: Training models that quantify fatigue, predict overtraining, or recommend personalized load adjustments based on biomechanical and physiological sensor fusion. - Injury Risk Prediction & Rehabilitation Monitoring: Building predictive models that identify musculoskeletal injury risk factors or track recovery trajectories from sensor-based movement assessments. - Sports Science Analytics: Developing AI-driven analysis tools for technique evaluation, energy expenditure estimation, or real-time performance feedback during training sessions. - Predictive Patient Monitoring: Training early-warning models for clinical deterioration, therapy non-response, or adverse events from continuous physiological data streams. - Synthetization and Transfer of AI Model to Edge Device: Re-training and quantizing an ML/DL/AI algorithm, so it can be run on an edge device (e.g. wearable), while optimizing performance (inference accuracy, energy consumption per inference, latency etc.) **Who We Are:** The **Fraunhofer Insitute for Integrated Circuits (Fraunhofer IIS)** has established the **"Center for Sensor Technology and Digital Medicine" (CEMDIS)** in cooperation with the Universitätsklinikum Erlangen and the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) to enhance modern healthcare through **innovative sensor technology** and **digital solutions**. This center focuses on integrating **innovative medical technologies** such as **wearables** and **robotic systems** to support **medical diagnostics**, **patient monitoring** and **evaluating patient-specific therapies** by providing digital health solutions für real-life healthcare. Located at the Universitätsklinikum Erlangen, it offers unique infrastructures for the **development**, **integration**, and **validation** of novel health technologies, providing companies opportunities for **technological advancements**. For more information, visit the [Fraunhofer IIS website](https://www.iis.fraunhofer.de/de/ff/sse/health/zentrum-sensorik-medizin.html).

Multitel (MULTITEL)

This service provides an independent and reproducible evaluation of AI model performance using well-established quantitative metrics. The evaluation is conducted in a controlled and documented execution environment to ensure traceability and repeatability of results. Performance is assessed on client-provided datasets and models, and associated measurement uncertainties are systematically analyzed and reported. The service delivers a detailed and interpretable evaluation report. All activities are performed under ISO 9001 certified processes.

Zilinska Univerzita V Ziline (UNIZA)

Tailored development and consultation of AI models, including supervised, unsupervised, and reinforcement learning approaches. This service supports SMEs in designing, training, validating, and deploying AI solutions. Technical details:Python-based frameworks (TensorFlow, PyTorch), model hosting options. Use cases/examples:SME is looking for information on how to evaluate their model or how to design adequate architecture.

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

Helping to rigorous testing and validation of the developed Multimodal AI models in real hospital environments based on real-world medical images and with collaboration with AI experts and medical staff. In collaboration with the medical staff and advanced deep learning algorithms, test the AI models on datasets not yet seen by the AI to confirm the correct behaviour of the AI. Discussion with medical experts and validation with new data sets (continuous learning) to ensure that the AI is still correct even with new data). Verification that there are no biases. Keywords: Medical Image Analysis; Deep Learning Algorithms; Binary Classification; Anomaly Localization; AI Model Development; Real-World Images; AI Testing and Validation; Collaboration with Medical Staff; AI Expertise; Custom Pretrained Pipelines; Clinical Imaging; Image Segmentation; Image Feature Extraction; Trustworthy AI

Fraunhofer Gesellschaft Zur Forderung Der Angewandten Forschung Ev (Fraunhofer)

This service supports the development, training, and systematic testing of AI and machine learning models in a controlled experimental environment. It enables technology providers to improve model performance, robustness, and generalization by combining structured training workflows with rigorous testing and evaluation procedures prior to deployment or further validation steps.   The service covers the supervised training and testing of AI/ML models using customer‑provided data and problem specifications. Depending on the use case, model training is performed using established machine learning and deep learning methods, followed by structured testing on independent datasets.
Testing and evaluation activities focus on assessing model performance, stability, and generalization behaviour under defined conditions. The process and results are documented in a technical report, supporting iterative model improvement and downstream validation or certification activities. The service applies state‑of‑the‑art methods from applied machine learning and experimental AI research, following best practices for data handling, model training, and performance evaluation as commonly used in academic and industrial AI development.   Depending on the task and model type, evaluation outcomes may include: • Quantitative performance metrics (e.g. accuracy, error measures, task‑specific scores) • Evidence of improved generalization on held‑out test data • Identification of performance limitations, failure modes, or data‑related issue This service is offered by Fraunhofer HHI.

Fraunhofer Gesellschaft Zur Forderung Der Angewandten Forschung Ev (Fraunhofer)

This service provides systematic validation and quality assessment of artificial intelligence and machine learning (AI/ML) models. The goal is to objectively assess model performance, robustness, and reliability under controlled and reproducible conditions, supporting trustworthy AI development and deployment. The provided AI/ML model is evaluated against a user‑supplied test dataset using established validation methodologies. Based on the assessment results, a comprehensive validation report is generated, summarizing quantitative performance metrics and key quality indicators. The validation process follows state‑of‑the‑art scientific approaches and ensures transparency and reproducibility of results. The service delivers a set of quantitative AI/ML performance metrics, which may include (depending on the use case and model type): • Model accuracy and error measures • Sensitivity, specificity, precision, recall, and related metrics • Additional task‑specific or domain‑relevant performance indicators

Laboratoire National De Metrologie Et D'Essais (LNE)

To ensure that their solution works accordingly regarding a task, AI provider have to evaluate their system using a specific dataset. The performance obtained on this dataset helps to prove the adequate performance of their AI solution. However, the evaluation process can be often quite hard for an AI provider to do properly: the evaluation dataset needs to be correctly qualified, and the creation of the evaluation protocol as well as the analysis of the results are not easy tasks. This service allows AI provider to benefit of the LNE expertise with a full evaluation of their AI system: using an evaluation dataset created for the test or provided by a partner of the TEF project, an assessment of the performance of the AI system is done, and an analysis of its behavior provided. This work also includes a quality assessment of the evaluation dataset. The scope of the analysis of the evaluation results can include robustness and resilience evaluation, depending of the needs of the SMEs. With this service, the customer will have a full assessment of its solution, allowing them to answer the accuracy requirements of the AI regulation, while also having a full report following all transparency and reproductibility requirements. This service generally takes around 2 months, depending of the needs of the customers.

Laboratoire National De Metrologie Et D'Essais (LNE)

The system provided by the customer, for instance a medical device using an AI module, will be tested using a simulated environment. The service will be provided using the LE.IA Simulation, a fully simulated test environment (the robot or the medical device and its operating environment are simulated). It allows to generate a very large number of test scenarios that are not feasible in a physical environment.

RISE Research Institutes Of Sweden

Method description: Preprocessing Data to be AI-ready. This service can be a preprocessing step to perform other AI services from RISE, for example RISE's service "AI improvement and development". The service is customized according to the SME's needs. Method reference: To be refined depending on the type of task, and in dialog with the customer. The Center for Applied AI at RISE carries out cutting-edge research in AI, connects expertise and applications within RISE, and explores a wide range of innovative applications with industry and the public sector. Applied AI Centre at RISE helps companies and government agencies to see more potential in the technology, use it more wisely and develop it faster. Keywords: Trustworthy AI, Federated learning, Data Science, AI, Deep Learning, Natural language processing, Computer Vision, Machine Learning.

Karolinska Institutet (KI)

## Overview This service provides a structured evaluation of AI development pipelines and technical documentation, tailored for SMEs and researchers in healthtech. Leveraging clinical and regulatory expertise at SMAILE, Karolinska Institutet, the assessment covers traceability, software versioning, model lifecycle management, and documentation practices. The process ensures that development pipelines meet regulatory expectations, reproducibility requirements, and are aligned with future audit or MDR/AI Act evaluations. We provide expertise and support in: - AI documentation & pipeline evaluation - MDR readiness & traceability compliance ### How can the service help you? This service addresses the lack of clear documentation and reproducible workflows in early-stage AI development. It gives you a clear roadmap for improvement and regulatory preparedness. Whether you're pursuing CE marking or preparing for due diligence from investors, this service boosts your confidence and credibility. ### How the service will be delivered? Delivered virtually by SMAILE experts at Karolinska Institutet. Includes documentation review and remote interviews. Typically executed over 1–3 weeks. --- ## Additional information ### Provider description The SMAILE core facility at Karolinska Institutet offers expert consulting and virtual testing for AI system validation, regulatory preparation, biomedical signal analysis, and data-driven life science solutions. SMAILE collaborates across KI, SciLifeLab, and RISE under the Swedish TEF-Health node. ### Technical description The service includes a structured audit of AI development stages using a checklist of regulatory, technical, and reproducibility indicators. The pipeline’s documentation is reviewed, highlighting gaps and providing targeted improvement suggestions. ### Service customization The scope can be tailored to specific pipeline stages (e.g., pre-processing, model training, evaluation). Clients may combine this service with model benchmarking or regulatory consultation.

Laboratoire National De Metrologie Et D'Essais (LNE)

AI resilience evaluation: evaluation of an AI system using a test dataset. This dataset will contain erroneous data designed to disturbed the behavior of the AI system. The general methodology will include the design of test protocols, the realization of tests, the analysis of results and production of a test reports.

Laboratoire National De Metrologie Et D'Essais (LNE)

Evaluation of an AI system using a test dataset augmented with artificial data. The new data results in a wide coverage of the operating range and addition of artificial noise based on the nature of the data processed by the AI system. The LNE data augmentation tools can apply physically-informed transformations to visual data and timeseries data. The visual transformations include for instance: gaussian (electrical) and poisson (thermal) noise, gaussian blur (focus issues), loss of pixels, lines and columns (CCD failures). The timeseries transformations include random gaussian, laplace or uniform noises.

HUMANI SC - Hospital network (HUMANI)

Deployment and evaluation of AI solutions (algorithms, applications, etc.) in a secure IT environment (servers, virtual machines, etc.) within the hospital. This service, provided by clinical and IT experts, includes a technical evaluation to ensure interoperability, accessibility, and interpretability with existing IT systems and a comprehensive evaluation within real-world clinical environments.

RISE Research Institutes Of Sweden

Method description: This service performs testing, evaluation and verification of AI systems in light of the EU AI-Act. Methods consists of traditional performance testing coupled with formal methods to prove safety and functionality requirements, if applicable. The requirements put on the models are formulated in collaboration with the customer. RISE performs the corresponding virtual tests and records the results in a test and verification report. Method reference: To be refined depending on the type of task, and in dialog with the customer. The Center for Applied AI at RISE carries out cutting-edge research in AI, connects expertise and applications within RISE, and explores a wide range of innovative applications with industry and the public sector. Applied AI Centre at RISE helps companies and government agencies to see more potential in the technology, use it more wisely and develop it faster. Keywords: AI systems, Data Science, Deep Learning, Natural language processing, Machine Learning, Trustworthy AI

Politecnico Di Milano (POLIMI)

Conceiving, developing and optimization of AI algorithmic approaches across the whole pathway: 1) Definition of applications objectives 2) Data collection, data curation and bias evaluation 3) Preprocessing and optimisation of feature extractions 4) Selection of suitable machine learning models, model training and testing 5) External validation 6) Interpretations of models (XAI algorithm applications and development). Keywords: AI, algorithm development, AI testing and validation, Data Curation, Feature Extraction, Learning Models, External Validation, XAI

Friedrich-Alexander-Universitaet Erlangen-Nuernberg (FAU)

# Overview This service delivers specialized technical consultation and machine learning expertise to optimize, refine, and enhance artificial intelligence algorithms for digital health applications. Building robust algorithms requires continuous iteration across model architectures, feature engineering, and training pipelines. Through tailored expert evaluation, the service supports startups and SMEs in identifying performance bottlenecks, improving model accuracy, and optimizing computational efficiency. The primary objective is to accelerate the technological readiness of health-focused algorithms by applying agile development practices and peer-reviewed data science methodologies to real-world medical data challenges. ## How can the service help you? Developing algorithms internally often leads to plateaued model performance, overfitting, or inefficiency when handling complex biomedical datasets. This service provides targeted external expertise to address these technical challenges directly. - *Before:* An algorithm hampered by sub-optimal accuracy, high compute requirements, or poor generalization across real-world data distributions. - *After:* An upgraded model architecture supported by refined training strategies, improved predictive performance, and robust validation protocols. It enables technical teams to solve complex data science roadblocks, speed up product iteration cycles, and ensure their core algorithms meet necessary technical standards. ## How will the service be delivered? The service is conducted through remote technical consultations, code/model reviews, and joint collaborative sessions. Client teams provide their specific requirements, performance targets, or current algorithmic pipelines. Our experts analyse the existing implementation and deliver tailored technical guidance and actionable optimization strategies. # Additional information ## Provider description Operating from the Department of Artificial Intelligence in Biomedical Engineering at Friedrich-Alexander-Universität Erlangen-Nürnberg, we are a service provider node within the TEF-Health consortium. The research group specializes in machine learning, biomedical signal processing, and multimodal sensor synchronization. Our team provides testing infrastructure and scientific support for evaluating medical devices, wearables, and contactless sensing systems. ## Technical details The advisory framework draws upon established agile development principles and modern deep learning / machine learning methodologies for biomedical data processing. - *Input Data Requirements:* Project requirements, performance metrics, model architecture documentation, or anonymized training data samples. - *Core Tasks & Processing:* Code and architecture audits, feature engineering assessment, hyperparameter optimization strategies, bias/overfitting analysis, and algorithmic benchmarking against medical data standards. - *Outputs:* Technical consultation reports, code optimization recommendations, and structured technical documentation detailing improved model pipelines. - *Scientific Reference:* Aligned with peer-reviewed research on biomedical algorithm development ([DOI: 10.1109/OJEMB.2024.3356791](https://ieeexplore.ieee.org/document/10411039)). ## Service customization The scope of engagement is flexible and customized to match the client's development stage. SMEs can choose focused guidance on specific tasks (e.g., feature selection or model hyperparameter tuning) or request a broad end-to-end algorithmic review across the full research and development lifecycle.

Laboratoire National De Metrologie Et D'Essais (LNE)

Study and assess the maturity of the AI system in regards of the requirements of the AI Act, especially for high risk AI systems (Articles 9 to Articles 15). The service consists of an analysis of the documentation of the AI system provider. In addition to the documentation, a meeting akin to an audit is organized on specific points to better understand the processes used in developping the AI system. A report is provided at the end of the analysis, which contain the observations of the LNE regarding the maturity of the AI system compared to the requirements of the AI Act.

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

Logistic & Process Optimisation: Provide expertise on how to address large scale logistic, scheduling, and other combinatorial optimisation problems using constraint-based local search. CETIC has developed a routing optimization algorithm called RoutaR. It is based on the oscar.cbls optimization engine and on GraphHopper cartography. Both are open source. Method Description: It inputs a routing problem, expressed in a Json file format. It typically contains a set of geographic locations, a description of a fleet of vehicles and a set of constraints. It produces a routing solution; a planning of each vehicle of the considered fleet that mentions their tasks and geographic locations Method reference: https://www.cetic.be/RoutaR-outil-de-planification-performant-et-facilement-adaptable

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

The collection of data from various sources (devices, files, interconnected services, etc.) is a complex step and typically falls outside the core business of medical sector companies, which aim to focus on the analysis of this data. How can the service help you? The CETIC provides access to a highly customizable catalog of tools for engineering medical data collection, storage and analysis. These tools enable : (a) automation of the collection of heterogeneous medical data from virtually any type of source (like IoT, API, files, remote repositories or databases, propriatary equipments, etc.) and its mapping to fully-customisable data representations enriched with semantics (DMWay tool) (b) a scalable and persistent data storage (c) easy yet powerful data analysis (TSANO tool) How the service will be delivered? It depends on the customer's need Method reference: https://asset.cetic.be/en/dmway/ https://www.cetic.be/analyse-prescriptive-au-service-de-industrie40

Karolinska Institutet (KI)

## Overview This service focuses on assessing the fairness and equity of AI models used in healthcare. Through demographic performance breakdowns, explainable AI (XAI) tools, and bias detection methods, SMAILE at Karolinska Institutet supports the ethical development of AI tools. The service identifies potential bias in models and helps adjust training datasets or algorithms to promote fairness across age, gender, ethnicity, and other protected attributes. We provide expertise and support in: • Bias and fairness audits for clinical AI • Performance disaggregation across demographic groups • Explainable AI and equity-by-design principles ### How can the service help you? Many developers are unaware of hidden biases in training data or model outputs. This service reveals fairness issues early and provides actionable diagnostics with tailored mitigation strategies. It helps build stakeholder trust and prepares for future ethical or regulatory scrutiny. ### How will the service be delivered? Primarily delivered virtually. Clients provide access to anonymized datasets and models. SMAILE experts run audits using stratified metrics and XAI tools. Turnaround time: 2–3 weeks. ## Additional Information ### Provider Description SMAILE specializes in responsible AI and ethical technology development. We support equity-focused model design and provide tools and strategies to mitigate risks of algorithmic harm in clinical settings. ### Technical Description Bias analysis is conducted using group-specific performance metrics (e.g., sensitivity, specificity, precision), fairness indicators (e.g., demographic parity, equal opportunity), and XAI tools such as SHAP, LIME, and counterfactual testing. The service aligns with fairness frameworks, including IEEE P7003 and the EU AI Act. ### Service Customization Clients can specify target subgroups for analysis. Optional components include bias mitigation training, fairness KPI monitoring, and strategy workshops.

Karolinska Institutet (KI)

## Overview This service supports the development and validation of biomedical signal processing pipelines for healthcare applications. SMAILE at Karolinska Institutet provides guidance for processing a range of biosignals, including ECG, EEG, EMG, respiration, accelerometers, gyroscope, and electrical bioimpedance. The focus is on signal filtering, feature extraction, event detection, and pipeline validation, aligned with clinical research standards. ### We provide expertise in: • Biomedical signal filtering and preprocessing • Feature extraction and clinical event annotation • Signal Analysis and Modeling • Quality assurance and reproducibility ### How can the service help you? Many early-stage companies and research projects lack robust, validated signal pipelines. This service enables reproducible analysis, clinical readiness, and regulatory alignment for biosignal-driven products and studies. ### How will the service be delivered? Can be delivered virtually or in conjunction with lab access. Clients share sample datasets, objectives, and processing goals. Duration: 2–6 weeks based on complexity. ## Additional Information ### Provider Description SMAILE provides advanced support in biomedical signal analytics, with experience in wearable health monitoring, digital phenotyping, and clinical-grade biosignal systems. ### Technical Description Pipelines are constructed and validated using Python (SciPy, MNE, NeuroKit), MATLAB, or other signal tools. ### Service Customization Clients may request focused support for real-time processing, offline batch analysis, or integration with AI modules. Output formats, sampling rates, and noise tolerance are adjustable.

Universita Degli Studi Di Pavia (UNIPV)

This service offers the construction and simulation of brain digital twins for neurological and neuropharmacological applications. The simulations are typically applicable to single subject data for developing personalized medicine and pharma. Method description: Raw structural and functional data is preprocessed and analyzed to create brain connectivity data. The computational framework is tuned against experimental data to simulate brain dynamics and optimize physiological parameters, e.g., excitatory and inhibitory activity. In order to obtain personalized digital brain twin single-subject data must be used. Method reference: Local method respecting General Data Protection Regulation (GDPR)

Virtual Research Environment / Health Data Cloud

## Overview The Charité Virtual Research Environment (VRE) provides a Secure Processing Environment (SPE) and Trusted Research Environment (TRE) for collaborative processing, exchange, and analysis of human health data in the context of GDPR, AI Act compliance, MDR evidence generation, and Post-Market Clinical Follow-Up (PMCF). The service enables controlled collaborative processing of sensitive health data within isolated and audit-capable project environments aligned with GDPR and institutional information security requirements. The VRE supports isolated project workspaces on shared multi-tenant infrastructure with secure remote access, managed analysis environments, and traceable processing workflows. ### VRE services include: - Secure isolated project workspaces - Secure data exchange between Charité and external partners - Trusted Research Environment / Secure Processing Environment and secure collaboration workflows for sensitive health data like imaging, genetic, or clinical data - Remote Desktop virtual machines - Managed Jupyter and Linux analysis environments - GPU-enabled compute resources - Long-term storage, backup, and recovery mechanisms - Audit logging and traceability support - Monitoring and operational oversight - Guacamole/VPN-based remote access - Support for automated pseudonymisation and anonymisation workflows - SOPs, onboarding, and user support - Project coordination and operational support - GDPR, MDR, and AI Act support processes - Collaborative data analysis and digital evidence generation --- ## How can the VRE help you? The VRE enables secure and GDPR-compliant collaborative research with sensitive human health data that cannot be fully anonymized, including imaging, genomics, clinical, and other highly sensitive biomedical datasets. The platform provides controlled and traceable access to protected data within isolated project environments designed for secure collaborative processing and analysis. The VRE supports interdisciplinary collaboration between research institutions, healthcare providers, and industrial partners while reducing the need for uncontrolled data transfers and decentralized data handling. Managed analysis environments, secure remote access mechanisms, audit logging, and controlled collaboration workflows support reproducible and secure data processing in compliance-sensitive research settings. The VRE can additionally support translational and regulatory-oriented activities such as AI system evaluation, PMCF activities, secure secondary use of health data, federated collaborations, and evidence generation for MDR and AI Act related processes. --- ## How is the VRE accessed? The VRE is primarily accessed through secure remote access mechanisms. Dedicated isolated project spaces are provisioned on Charité-operated infrastructure. Depending on project scope, the service may range from lightweight onboarding and workspace provisioning to fully managed collaborative environments with operational and compliance support. Typical project durations range from several weeks to multiple years. --- ## Provider Description Charité operates advanced research and digital infrastructure environments for secure biomedical and clinical data processing in research and translational medicine contexts. The VRE supports interdisciplinary collaboration between clinical, scientific, and industrial stakeholders in compliance-sensitive environments. --- ## Technical Description The platform is based on shared multi-tenant infrastructure with isolated project environments. Available resources include virtual machines, GPU-enabled compute nodes, storage systems, managed analysis environments, backup systems, authentication and authorization services, and monitoring capabilities. The environment supports secure remote access, auditability, traceability, and controlled data exchange workflows for sensitive health data processing. --- ## Service Customization The service can be customized depending on project requirements, including: - Dedicated or shared compute environments - GPU-enabled processing - Long-term storage requirements - Custom virtual machine sizing - Federated collaboration scenarios - Compliance and documentation support - Enhanced operational support models - Extended onboarding and training - Integration with external infrastructures and workflows

HUMANI SC - Hospital network (HUMANI)

HUmani cleans and annotates clinical datasets from external providers.

HUMANI SC - Hospital network (HUMANI)

Assessment of AI solutions within real-world clinical environments by healthcare and IT professionals. This service covers the entire testing lifecycle: from formative assessments (iterative clinician feedback to refine the product and its integration) to summative validation (structured performance testing against predefined KPIs). It ensures AI solutions are safety and clinically accurate, and optimized for user adoption, workflow efficiency and security.

Zilinska Univerzita V Ziline (UNIZA)

Development of intelligent, conversational chatbots powered by large language models (LLMs), tailored to specific healthcare, research, or business needs. These chatbots can understand and generate natural language responses, integrate with organizational knowledge bases, and support advanced tasks such as appointment scheduling, information retrieval, or interactive education. Technical details:Python, LLMs, custom datasets. Use cases/examples: SME wants to develop a chatbot service that would help users operate with their AI system.

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

Providing prebuilt AI pipelines tailored for medical image analysis. Streamlining AI model integration into existing healthcare workflows. Accelerating the deployment of AI solutions in clinical settings. Using MLOps and DVC to accelerate the AI development and deployment with CI-CD pipelines. Keywords: Medical Image Analysis; Deep Learning Algorithms; Binary Classification; Anomaly Localization; AI Model Development; Real-World Images; AI Testing and Validation; Collaboration with Medical Staff; AI Expertise; Custom Pretrained Pipelines; Clinical Imaging; Image Segmentation; Image Feature Extraction; Trustworthy AI

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

Cyber-Physical Systems Cybersecurity Testing ensures cyber-physical systems (CPS), such as robotic systems, comply with security requirements and standards (such as Cyber Resilience Act). Security testing includes security functional testing of the whole system, and vulnerability scanning of the software components. The overall goal is to release your product without known important vulnerabilities. How can the service help you? Our service is based on a custom platform that automates vulnerability scanning, penetration testing and security functional testing, with the possibility to integrate our platform in your DevSecOps activities for full security activities automation. The focus on cyber-physical systems takes the form of support for physical interfaces and buses/protocols, with possible attacks on RF signals (GPS or WiFi), for example. Our offer differentiates from existing ones with a unique risk-based approach, where a security risk analysis drives the whole process of security testing, and by the use of open-source tools to avoid vendor lock-in. The method is based on existing standards : PTES, Etsi, NIST, FDAM... How the service will be delivered? Based on information provided by the customer about their system, this service offers to use our Automated Cybersecurity Testing platform, tools and method to define and perform cybersecurity tests, either on site or in our lab. Optionally, we can integrate our platform in your DevSecOps chain for full automation of security activities. Optionally, we can perform the security risk analysis that is used as input for the whole process. We will provide complete test reports as well as recommendations to lower the residual risk level and attack surface of your system. Together with the risk analysis, they can be used as a complete set of evidences towards authorities and customers. Service deployment: The service is deployed either on site or in our lab. Our test platform is made of several components : a server that stores all relevant information and generate the reports, and test workstations that can be used in our lab or at customer sites directly. Service standards: PTES, Etsi, NIST, FDAM...

Karolinska Institutet (KI)

## Overview This service provides a complete organisation or re-organisation of your datasets as well as guidance in selecting the most appropriate metadata management methodology. We offer data cleaning (e.g., renaming, annotation, formatting, and duplicate removal), quality control, and anonymization tailored to SME needs. This service aims to provide assistance based on a dataset management plan. We follow community data and metadata standards in place and offer curated, up-to-date guidance based on FAIR principles (Findable, Accessible, Interoperable and Reusable) whenever possible. We have expertise in data and metadata standards in many life science fields, including, more specifically, imaging (PET, MRI, MEG, EEG) and structural data, genomics, metabolomics, and proteomics data. Expertise in legal and ethical aspects of dataset management can also be provided. We provide expertise and technical support in the following areas: - Project design & management - Legal & ethical support - Data cleaning, annotation, anonymisation ### How can the service help you? This service helps SMEs improve the organization and usability of their datasets by providing data cleaning, quality control, and anonymization tailored to their needs. With expert guidance on metadata management and adherence to FAIR principles, it ensures your data is reliable, compliant, and ready for effective use, building trust and confidence in your data-driven initiatives. ### How the service will be delivered? The service will be delivered according to established ethical agreements and guidelines, in collaboration with the SME and researchers from the Swedish TEF-Health node. Data processing is facilitated through a secure virtual environment managed by Karolinska Institutet, ensuring the highest standards of data protection. --- ## Additional information ### Provider description The Swedish TEF-Health node is a collaboration between Karolinska Institutet, SciLifeLab and RISE, and is led by Karolinska Institutet. Together, we offer world-leading services with our unique collection of core facilities. We can grant services in expert consulting, virtual- and physical testing in the range of in vivo imaging, ex vivo OMICS, pharmaceutical development, simulated healthcare environments, AI-system validation and development, advanced data analysis and other data-driven life science. ### Technical description The service restructures and processes datasets within a secure computing environment utilizing advanced data curation and metadata management frameworks. Data preprocessing operations, including nomenclature standardization, semantic annotation and format normalization are executed in compliance with domain-specific ontologies and if possible FAIR principles. Quality assurance protocols are implemented to validate data integrity and compliance. Metadata schemas are optimized for interoperability and reuse, leveraging established standards such as RDF and JSON-LD. ### Service customization The service can be customized according to your specific needs. It may be required to combine this service with other services on offer. --- ## Use case example ### Context A life science SME specializing in metabolomics has developed a novel pipeline for identifying biomarkers in rare metabolic disorders. They have generated extensive LC-MS/MS datasets from patient samples across multiple studies. However, the datasets are stored in a mix of proprietary formats, lack harmonized metadata and do not meet the submission requirements of repositories like MetaboLights. This situation delays the publication of their findings, limiting their visibility and ability to secure collaborations or funding for further pipeline validation. ### Objective To clean, harmonize and annotate the SME’s LC-MS/MS datasets according to MetaboLights requirements. Ensure compliance with FAIR principles to facilitate immediate repository submission and support future scalability. ### Solution The SME engages with the Swedish TEF-Health node to reorganize and optimize their metabolomics datasets for repository submission. Experts provide tailored data cleaning, format conversion, and metadata harmonization, ensuring compatibility with repository standards and enabling wider reuse of the data. ### Implementation #### Ethical Agreement The SME enters into an ethical agreement with researchers from the Swedish TEF-Health node, ensuring all data collection and usage complies with GDPR and national Swedish regulations. #### Secure Access Usage of the collected data is facilitated through a secure virtual environment managed by Karolinska Institutet, ensuring the highest standards of data protection. #### Data Processing and Format Conversion The SME’s raw datasets, stored in various proprietary formats, are converted into open formats such as mzML and mzTab, which are compatible with repository requirements. During this process, quality control measures, including noise filtering and peak annotation, are applied to improve data integrity and reliability. #### Metadata Harmonization Metadata schemas based on MetaboLights standards are created. These schemas incorporate essential details about study design, sample preparation and instrument parameters. Ontology based annotation is applied to harmonize metadata across all datasets, ensuring consistency and compliance with repository guidelines. ### Benefits - **FAIR Data Management**: Enhances SMEs’ ability to manage and share clinical trial data effectively while ensuring interoperability and quality. - **Stakeholder Credibility**: A well-structured data management plan builds trust with regulators, funders, and collaborators. - **Regulatory Compliance**: Ensures adherence to ethical and legal standards, reducing data handling risks. - **Data Sustainability**: Supports long-term usability and scalability, enabling future research opportunities. ### Impact The SME’s biomarker discovery pipeline gains recognition as a reliable and validated tool in the rare disease research community. Their repository-submitted data fosters new collaborations with academic researchers and industry stakeholders, accelerating the translation of their findings into clinical applications.

RISE Research Institutes Of Sweden

Method description: Integration of data from different datasets. Not exclusively, however this service can be performed as a data preprocessing step to "AI improvement and development" service available from RISE. The service is customized according to the SME's needs. Add-on service to other services. Method reference: To be refined depending on the type of task, and in dialog with the customer. The Center for Applied AI at RISE carries out cutting-edge research in AI, connects expertise and applications within RISE, and explores a wide range of innovative applications with industry and the public sector. Applied AI Centre at RISE helps companies and government agencies to see more potential in the technology, use it more wisely and develop it faster. Keywords: Trustworthy AI, Federated learning, Data Science, AI, Deep Learning, Natural language processing, Computer Vision, Machine Learning.

RISE Research Institutes Of Sweden

Method description: Analysis of big datasets. Not exclusively, however this service can be performed as a data preprocessing step to "AI improvement and development" service available from RISE. The service is customized according to the SME's needs. Add-on service to other services. Method reference: To be refined depending on the type of task, and in dialog with the customer. The Center for Applied AI at RISE carries out cutting-edge research in AI, connects expertise and applications within RISE, and explores a wide range of innovative applications with industry and the public sector. Applied AI Centre at RISE helps companies and government agencies to see more potential in the technology, use it more wisely and develop it faster. Keywords: Trustworthy AI, Federated learning, Data Science, AI, Deep Learning, Natural language processing, Computer Vision, Machine learning.

Fondazione Bruno Kessler (FBK)

This service covers all aspects regarding the analysis of data useful to understand the task, including problem identification, data curation and preprocessing, statistical and exploratory data analysis, interpretation and visualization. Keywords: Biomedical data Analysis, Data Processing, Data Science.

Centre Hospitalier Universitaire De Rennes (CHU RENNES)

This service provides access to health data from Clinical Data Warehouses of CHU Rennes or/and CHU Grenoble-Alpes and from other clinical data warehouse networks. It includes datamart building (cohort and document selection, mapping to standard...) and data qualification (extraction of clinical variables, data cleaning, clinical consistency checking...) steps. In this way, the service offers companies an opportunity to access healthcare data and/or descriptive statistical analyses.​ This service involves only structured data extraction and pre-processing. In cases where the data is unstructured (e.g., requiring natural language processing for text, feature extraction from images, etc.), a specialized algorithm development service will be required (see our service "Algorithm development" for more details).​ Prerequisite : To have data access, it is essential to have CNIL (Commission Nationale de l'Informatique et des Libertés) approval, a favorable opinion from the Scientific and Ethics Council (CSE), and a signed contract between SME and hospital. Furthermore, our services "Feasibility study" and "Methodological advising" must have been previously conducted.​ Method Description: The service includes a minimal mandatory task set. Optional tasks can be jointly selected depending on the SME needs and available resources: - Datamart building - Patient selection (cohort) [mandatory] - Selection of data of interest [mandatory] - Mapping to FHIR/OMOP standards [optional] - Data qualification steps - Extraction of variables of interest [optional] - Data cleaning [mandatory] - Clinical consistency check by expert [mandatory] - Annotation/segmentation [optional] - Descriptive statistical study [optional] - Confidentiality control [mandatory] - Matching with other databases (i.e., SNDS) [optional] - Deployment of methods on another warehouse [optional] - Datamart availability on Hospital platform (CHU platform) [Standard offer] Method reference: Based on a scientific reference or a method designed by CHU Rennes and its partners (LTSI, LaTIM, etc.), as well as on methodologies provided by WP6. Provision by respecting General Data Protection Regulation (GDPR), AI Act, Good Clinical Practice (GCP) and french legal framework

Centre Hospitalier Universitaire De Rennes (CHU RENNES)

This service tends to offer a scalable platform to access dataset by the SME taking into account its infrastructure requirements and french legal framework.​ Prerequisite : The dataset must exclusively originate from our data warehouse. Method Description: Depending on the SME needs and available resources, this service may include all or part of the following task list: - Deployment of the dataset on an environment outside CDW Method reference: Provision by respecting General Data Protection Regulation (GDPR), AI Act, Good Clinical Practice (GCP) and french legal framework

Centre Hospitalier Universitaire De Rennes (CHU RENNES)

This service offers method development for tasks at different processing levels (pre-processing, feature extraction, statistical, machine learning and/or deep learning algorithms). In short, it offers support for the full exploitation of data at all processing stages. With this service, companies can benefit from clinical and data science expertise to facilitate the entire healthcare data processing process, including raw data cleaning, extraction of relevant features, statistical analysis and implementation of advanced algorithms. Specific capabilities include text processing using regular expression (regex) filtering or NLP techniques, feature extraction from images, physiological signal through both Machine Learning and Deep Learning algorithms.​ Prerequisite : The dataset must exclusively originate from our data warehouse. Our service "Data provision and analysis" must simultaneously be conducted. Method Description: The service includes: - Drafting of functional specifications [mandatory] - Development of data exploitation methods [mandatory] Method reference: Based on scientific reference or methods designed by CHU Rennes and its partners (LTSI, LaTIM, etc.), as well as on methodologies provided by WP6. Implemented methods will integrate General Data Protection Regulation (GDPR), AI Act, Good Clinical Practice (GCP) insights.

Multitel (MULTITEL)

This service provides a statistical analysis of datasets and dataset splits used in machine learning pipelines, with the objective of verifying their distributional consistency. It supports the validation of training, validation, and test splits by detecting statistically significant differences that could bias model training or invalidate performance evaluation. The analysis conducted in a controlled and documented execution environment to ensure traceability and repeatability of results. The service is particularly relevant for AI systems in the health domain, where dataset representativeness directly affects evaluation reliability, which is critical for regulatory compliance and clinical trustworthiness. The analysis is performed using robust statistical techniques and is delivered as a structured, interpretable report. The service execution follows ISO 9001 certified processes.

Centre Hospitalier Universitaire De Grenoble (CHUGA)

This service allows to validate the clinical question, the feasibility of SME demand. To be more specific, it first includes a validation of the clinical question with a clinical expert. Then, a pre-screening is conducted to assess data availability, leveraging the expertise of our data scientists, which leads to the production of a feasibility report. Prerequisite: If the need remains to be precised, our service "Initial Consultation for SMEs" must be conducted before starting the feasibility study.​ Method Description: The service includes :     Pre-screening [mandatory]     Feasibility study [mandatory] Method reference: Local method respecting General Data Protection Regulation (GDPR), AI Act and Good Clinical Practice (GCP)

Centre Hospitalier Universitaire De Grenoble (CHUGA)

This service provides access to health data from Clinical Data Warehouses of CHU Grenoble-Alpes. It includes datamart building (cohort and document selection, mapping to standard...) and data qualification (extraction of clinical variables, data cleaning, clinical consistency checking...) steps. In this way, the service offers companies an opportunity to access healthcare data and/or descriptive statistical analyses.​ This service involves only structured data extraction and pre-processing. In cases where the data is unstructured (e.g., requiring natural language processing for text, feature extraction from images, etc.), a specialized algorithm development service will be required (see our service "Algorithm development" for more details).​ An extension of the study to other clinical data warehouses from the TEF consortium can be considered in a second time. Prerequisite : To have data access, it is mandatory to fullfil the regulatory package asked by the French regulatory authorities* and a signed contract between the SME and the hospital. Furthermore, our services "Feasability study" and "Methodological advising" must have been previously conducted.​ Method Description: The service includes a minimal mandatory task set. Optional tasks can be jointly selected depending on the SME needs and available ressources: - Datamart building - Patient selection (cohort) [mandatory] - Selection of data of interest [mandatory] - Data qualification steps - Extraction of variables of interest [optional] - Data cleaning [optional] - Clinical consistency check by expert [mandatory] - Annotation/segmentation [optional] - Descriptive statistical study [optional] - Confidentiality control [mandatory] - Matching with other databases (i.e., SNDS) [optional] - Deployment of methods on another warehouse [optional] - Datamart availability on Hospital platform (CHU platform) [Standard offer] Method reference: Based on a scientific reference or a method designed by CHU Grenoble Alpes, as well as on methodologies provided by WP6. Provision by respecting General Data Protection Regulation (GDPR), AI Act, Good Clinical Practice (GCP) and french legal framework *Regulatory package: - Approval from the CNIL (Commission Nationale de l'Informatique et des Libertés) - Favorable scientific opinion (provided internally in the hospital by a scientific committee) - Favorable ethical assessment (provided internally in the hospital by an Institutional Review Board (IRB))

Centre Hospitalier Universitaire De Grenoble (CHUGA)

This service tend to offer a scalable platform to access dataset by the SME taking into account its infrastructure requirements and french legal framework.​ Prerequisite : The dataset must exclusively originate from our data warehouse. Method Description: Depending on the SME needs and available ressources, this service may include all or part of the following task list: - Deployment of the dataset on an environment outside Clinical data Warehouse Method reference: Provision by respecting General Data Protection Regulation (GDPR), AI Act, Good Clinical Practice (GCP) and French legal framework

Centre Hospitalier Universitaire De Grenoble (CHUGA)

This service offers method development for tasks at different processing levels . With this service, companies can benefit from clinical and data science expertise to facilitate the entire healthcare data processing process, including raw data cleaning, extraction of relevant features (pre-processing), statistical analysis and implementation of advanced algorithms in some cases. In particular, we provide specific capabilities including text processing using NLP techniques, feature extraction from images through both Machine Learning and Deep Learning algorithms.​ Prerequisite : The dataset must exclusively originate from our data warehouse. Our service "Data provision and analysis" must simultaneously be conducted. Method Description: The service includes: - Drafting of functional specifications [mandatory] - Development of data exploitation methods [mandatory] Method reference: Implemented methods will integrate General Data Protection Regulation (GDPR), AI Act, Good Clinical Practice (GCP) insights.

Centre Hospitalier Universitaire De Grenoble (CHUGA)

This service will offer the possibility for SME to navigate interactively throught datasets thanks to an API developed within the CHUGA. It allows to select interactivly data and to visualize it thanks to BI tools. A short period of training is needed to be fully autonomous and able to exploit the full potential of the app. Method Description: Deployment of an API on a secure environment to navigate and vizualize interactively data.

Centro Hospitalar De Sao Joao Epe (CHSJ)

The service offers SMEs access to high-quality, annotated clinical datasets with relevant clinical for testing purposes, also for re-train algorithm models. These annotated datasets enable SMEs to validate their products, such as machine learning algorithms or medical devices, using real-world clinical data.

Unidade Local De Saúde De Coimbra EPE (ULS Coimbra EPE)

The service provides access to high-quality, annotated clinical datasets that have been curated and enriched to ensure relevance, quality, and suitability for healthcare research, development, and innovation, that are relevant for testing purposes and also for retraining algorithm models. These annotated datasets enable SMEs to validate their products, such as machine learning algorithms or medical devices, using real-world clinical data.

Centro Hospitalar De Sao Joao Epe (CHSJ)

This service offers SMEs access to curated clinical datasets along with expert guidance in selecting relevant variables for their specific use cases. Levaraging the hospital's extensive data resources and by thoroughly understanding the clinical requirements and objectives of each study, it ensures that the chosen variables are relevant, high-quality, and fit for purpose, thereby enabling more robust, efficient, and clinically meaningful analyses. This service assists SMEs in identifying the most pertinent variables from clinical datasets to support their product development efforts.

Unidade Local De Saúde De Coimbra EPE (ULS Coimbra EPE)

This service offers SMEs access to curated clinical datasets along with expert guidance in selecting relevant variables for their specific use cases. Levaraging the hospital's extensive data resources and by thoroughly understanding the clinical requirements and objectives of each study, it ensures that the chosen variables are relevant, high-quality, and fit for purpose, thereby enabling more robust, efficient, and clinically meaningful analyses. This service assists SMEs in identifying the most pertinent variables from clinical datasets to support their product development efforts.

Spms - Shared Services Of The Ministry Of Health, E.P.E

This service offers SMEs access to AI test datasets based on defined criteria, existing legislation and models E.g. REST and SOAP WEB Services Method Description: Analysis of the data requested including feasibility, ethical requirements, meeting with the company, data extraction and validation and secure access. Method reference: Data privacy impact assessment (https://www.cnil.fr/en/privacy-impact-assessment-pia), Regulation (EU) 2016/679 GDPR, Portuguese Law No.58/2019

Laboratoire National De Metrologie Et D'Essais (LNE)

With the development of AI models in most sectors of society, the importance of the data used to train and test these models has a been an important concern for the AI supplier. Understanding the quality of the datasets used, especially in sector such as the medical domain where data can be hard to get, is thus a prerequesite to ensure the correct development of AI models and building trust regarding these approach for society. This service aims to answer this challenge by assessing the quality of datasets provided by the client. The features of the datasets that will be analyzed are based on agreed definition by the AI community: completeness, balance, diversity, accuracy... The service will help the client that want to ensure the quality of a dataset and understand the features of the data, thus answering the requirements regarding data quality of the AI regulation. A report will be provided to the customer at the end of the service, explaining the methodology and detailling all the analysis and conclusions of the study. The service generally take around one month to be provided once the data has been delivered to LNE. Work is generaly done directly on LNE infrastructure, but in case of confidentiality requirements, it is possible to adapt the work.

Laboratoire National De Metrologie Et D'Essais (LNE)

Development of an evaluation method for the AI solution: given the specifications and context of application of the AI system, the test will provide the evaluation plan to assess the performance of the system, in order for the customer to realize its own evaluation. This service will be based on the expertise of the LNE in creating evaluation plan given an AI system. The evaluation plan will contain the specification of the evaluation tasks, the creation of the test data, the metrics, and the evaluation protocol. All these elements will be used by the customers to conduct the evaluation of the AI system.

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

Combinatorial optimization solution for processes, scheduling and logistics optimization. Provide expertise on how to address large scale routing, scheduling, logistics and other combinatorial optimisation problems using constraint-based local search. Keywords: Resource Planning, Patient Transportation, Process Optimasation

Fraunhofer Gesellschaft Zur Forderung Der Angewandten Forschung Ev (Fraunhofer)

**Platform:** [DPM-Research](https://www.iis.fraunhofer.de/de/ff/sse/health/mobile-health-lab/managementsystem-klinische-studien.html): A customizable eCRF platform that streamlines data and patient management and documentation in clinical studies, aligned with international standards and data protection regulations. **Who Can Benefit:** - Clinical Researchers: For streamlined data management processes that support robust research methodologies and regulatory compliance. - Regulatory Compliance Specialists: To ensure that clinical studies adhere to required international and local regulations, minimizing risk and enhancing credibility. - Healthcare Organizations: To manage clinical data effectively, improving research outcomes and operational efficiencies. **Key Features:** - Customized Data Management Solution: Provides a tailored data management and documentation platform designed to support clinical research studies. - eCRF Platform: Offers an electronic Case Report Form (eCRF) system customized for collecting and managing clinical research data efficiently. - Compliance with Standards: Ensures all studies are conducted in compliance with ISO 14155 and GDPR (DS-GVO), guaranteeing high standards in clinical investigation and data protection. **Possible Applications:** - Clinical Studies: Supports the management of clinical trial and patient data, ensuring accuracy, integrity, and regulatory compliance throughout the research lifecycle. - Data Documentation: Facilitates comprehensive data collection and documentation, improving the reliability and reproducibility of clinical research findings. **Who we are:** The **Fraunhofer Insitute for Integrated Circuits (Fraunhofer IIS)** has established the **"Center for Sensor Technology and Digital Medicine" (CEMDIS)** in cooperation with the Universitätsklinikum Erlangen and the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) to enhance modern healthcare through **innovative sensor technology** and **digital solutions**. This center focuses on integrating **innovative medical technologies** such as **wearables** and **robotic systems** to support **medical diagnostics**, **patient monitoring** and **evaluating patient-specific therapies** by providing digital health solutions für real-life healthcare. Located at the Universitätsklinikum Erlangen, it offers unique infrastructures for the **development**, **integration**, and **validation** of novel health technologies, providing companies opportunities for **technological advancements**. For more information, visit the [Fraunhofer IIS website](https://www.iis.fraunhofer.de/de/ff/sse/health/zentrum-sensorik-medizin.html).

Karolinska Institutet (KI)

This service is a follow up to the Data Feasibility Study, and can only be delivered after that service has been completed and a potential data set that meets the requirements has been identified. Note that an additional fee, not covered by the call's price reduction, may apply. ## Overview Acquiring existing data sets—full-service data management. This service makes data sets identified in the Data Feasibility Study available to SME’s for TEF-project purposes, such as validation and testing of AI systems. Data can be collected from routine health records and local, regional, and national databases in collaboration with healthcare providers in Sweden. We provide expertise and technical support in the following areas: • Project design & management • Ethical application support • Data assembly • Data cleaning, annotation, anonymisation or pseudonymisation ### How can the service help you? Access to existing health data can help you validate your AI solution without needing to generate new data. We provide access to high quality datasets for validation and testing of your AI system, to facilitate placing it on the market and to increase its market readiness level, for example by supporting regulatory approval and building stakeholder/healthcare-provider confidence. Validation and testing data can be used for independent performance evaluation of your AI system in a real-world or simulated environment. Validation and testing without needing to generate new data frees up time to focus on model optimization and deployment. ### How will the service be delivered? The service will be delivered according to established ethical agreements and guidelines, in collaboration with the SME and researchers from the Swedish TEF-Health node. Data usage is facilitated through a secure virtual environment managed by Karolinska Institutet, ensuring the highest standards of data protection. --- ## Additional information ### Provider description The Swedish TEF-Health node is a collaboration between Karolinska Institutet, SciLifeLab and RISE, and is led by Karolinska Institutet. Together, we offer world-leading services with our unique collection of core facilities. We grant services in expert consulting, virtual- and physical testing in the range of in vivo imaging, ex vivo OMICS, pharmaceutical development, simulated healthcare environments, AI-system validation and development, advanced data analysis and other data-driven life science. ### Technical description Datasets identified in the Data Feasibility service will be made available to suit your AI system, supporting the testing and validating of your solution. The data will be tailored to your specific request to ensure it is compatible with your AI system and its real-world use, and with full support from Karolinska Institutet to ensure compliance with all relevant regulations, including GDPR and national Swedish regulations. Upon completion of this service, you will receive an evaluation of the performance of your AI system. ### Service customization The service can be customised according to your specific needs, taking into account which type of data you require. --- ## Use case example An SME developing an AI-based diagnostics system requests access to the retrospective datasets identified in the Data Feasibility Study service to independently validate the solution. Under an ethical agreement with the Swedish TEF-Health node, the service assembles and prepares the data and provides access through a secure virtual environment at Karolinska Institutet, where the system is tested in simulated and real-world scenarios. The resulting performance evaluation demonstrates the system's accuracy and reliability, supporting the SME's regulatory approval process and market entry.

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

Challenge : The client may request from a software solution provider that a copy of the source code be held by a neutral third party. In the case of AI solutions, training data may also be involved. Security and confidentiality are obviously important. Our strength : CETIC routinely offers a secure procedure for retaining the source code of software and can extend this to AI-based software by managing and preserving the datasets used, thus demonstrating the ability to reproduce the protected software. This service enables a procedure to be put in place to preserve the source code of AI-based software by a trusted third party. This service will manage and preserve the datasets used to develop AI models. Keywords: Software Escrow services

Eucaim

Cancer Image Europe provides a robust, trustworthy platform for researchers, clinicians, and innovators to access diverse cancer images, enabling the benchmarking, testing, and piloting of AI-driven technologies. EUCAIM Platform integrates a dashboard, a catalogue and a federated searching environment to discover Medical Imaging data related to cancer from a distributed federation of data holders. EUCAIM provides in some nodes processing capacity to securely access the data. EUCAIM is mainly intended for Data scientists to develop, validate or improve AI models or image postprocessing tools. The platform is accessible in https://dashboard.eucaim.cancerimage.eu/

HUMANI SC - Hospital network (HUMANI)

Access to annotated and curated HUmani's patient datasets along with expert guidance in selecting relevant variables for specific use cases.

Fraunhofer Gesellschaft Zur Forderung Der Angewandten Forschung Ev (Fraunhofer)

This module evaluates whether an AI model's decisions are based on meaningful, domain-relevant features or whether the model relies on undesired shortcuts and spurious correlations. The customer-provided AI model is analysed on a supplied test dataset using established explainable AI techniques, including attribution methods (e.g., LRP, CRP), concept-based explanations, and feature importance analysis. The evaluation identifies which input features drive model predictions and assesses their alignment with domain knowledge. Outcomes can be: Positive Indicators • Alignment of model explanations with established domain knowledge • Transparent and interpretable decision patterns • Consistent reliance on clinically/scientifically relevant features Negative Indicators • Detection of spurious correlations or shortcut learning (e.g., “Clever Hans” behaviour) • Reliance on irrelevant, confounding, or artefactual input features • Inconsistent explanation patterns across similar cases

Politecnico Di Milano (POLIMI)

This service provides recommendations on how to implement the explainability models on the AI system of interest. The implementation can also be performed by POLIMI expert. Keyword: Explainaibility of AI algorithm

Health Data Hub (Plateforme des Données de Santé)

This service provides in-depth data analysis expertise, specifically focusing on the exploitation of the French National Health Data System (SNDS) and, if applicable, its linkage with other datasets. Deliverables include commented R or Python programs for data processing and a detailed report presenting descriptive and analytical results, clear visualizations, conclusions, and recommendations, all to be completed within 60 days.

RISE Research Institutes Of Sweden

Federated Learning (FL) is a transformative approach to Artificial Intelligence that allows models to learn from data without that data ever leaving its original location. In the healthcare sector, where patient privacy and data security are paramount, this service enables organizations to collaborate and train powerful AI models across multiple hospitals or clinics while keeping sensitive records safely behind their own firewalls. This service provides a comprehensive framework for designing, prototyping, and implementing distributed AI solutions. Many high-impact healthcare applications are currently stalled because moving data to a central server is either technically impossible due to bandwidth or legally restricted by GDPR and the AI Act. Our service bridges this gap by bringing the "learning" to the data. Technical Note: Federated Learning is used exclusively during the training phase. Once training is complete, the resulting AI model is a standard, standalone tool that can be deployed and used at any location at any time, identically to a traditionally trained model. We work closely with you to investigate the feasibility of federated learning for your specific datasets, ensuring that the resulting AI models are both high-performing and privacy-compliant. As this service is closely linked to our Privacy-Preserving Methods and Implementation offering, we integrate advanced techniques such as differential privacy or secure multi-party computation to ensure that even the model updates themselves do not leak sensitive information. How can the service help you? • Unlock Siloed Data: Train AI on distributed datasets that cannot be moved or merged due to legal (GDPR) or technical constraints. • Enhance Privacy: Minimize data exposure by keeping raw patient records on-premises. • Collaborative Innovation: Enable multiple SMEs or healthcare providers to co-develop a robust "Global Model" that performs better than any "Local Model" trained in isolation. • Regulatory Compliance: Align your AI development with the strict requirements of the AI Act and healthcare data regulations. How the service will be delivered Logistics: The service is delivered through the RISE Applied AI Center and can utilize the V-Platform (a Trusted Research Environment). It can be configured for on-site deployment within your infrastructure or via a cloud-federated setup. Delivery period: Engagements are planned in phases, starting with a feasibility study and moving into prototyping and implementation. Duration: Typically ranges from 3 to 6 months, depending on the complexity of the data architecture and the AI model requirements. Customer requirement: The customer must provide access to relevant use cases, problem statements, and data documentation. While raw data stays with the customer, technical access for the "FL-client" software is required. Deliveries: Prototyping of the FL architecture, integration of privacy-preserving hooks, and technical documentation. Output: A functional federated learning pipeline and a refined AI model optimized for distributed environments. This service is highly modular and can be customized to match your specific industry standards, technical infrastructure, and the specific sensitivity level of your health data.

Politecnico Di Milano (POLIMI)

Generation of synthetic virtual cohorts leveraging statistical and data-driven methods and featuring realistic anatomies, flow fields, and boundary conditions. These virtual populations can serve multiple purposes, from in silico trials for testing new devices and procedures, to training large deep learning models for disease-specific tasks and discoveries. Method Description: The service provides synthetic datasets for AI algorithm development, AI algorithm testing and digital twin based simulations (fluidynamic/structural simulations, fluid structure interaction) Method reference: https://doi.org/10.1016/j.cmpb.2023.107468

Zilinska Univerzita V Ziline (UNIZA)

Provision of high-performance computing infrastructure powered by NVIDIA GPUs, enabling SMEs and researchers to train and test AI models in a scalable and secure environment. This service supports deep learning, computer vision, and large dataset processing tasks. Technical details:NVIDIA GPUs, remote access, Docker support, secured environment. Use cases/examples: SMEs without adequate training machines can rent full power of the high-end computers. The computer operates a platform for data storage, model training and evaluation.

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

We specialize in developing and evaluating medical image analysis models using deep learning algorithms for tasks such as binary classification and anomaly localization. Our AI models are rigorously tested and validated in real-world hospital settings through prospective studies, ensuring their accuracy and reliability. How can the service help you? We aim to improve the clinical relevance and effectiveness of our AI models by leveraging our experience of real-world datasets achieved through close collaboration with healthcare professionals. This hands-on approach allows us to develop and evaluate AI models in real-world contexts, ensuring that they meet the specific needs of healthcare applications and work reliably in clinical environments. How will the service be delivered? The service will be delivered in several steps: - Consultation & Needs Assessment : Understanding your specific requirements and use cases. - Model Development & Optimization : Designing and optimizing AI models for medical imaging tasks. - Testing & Validation : Testing the models with real-world datasets to ensure performance and reliability. - Pretrained Pipeline Delivery : Providing custom pretrained pipelines for your needs. - Ongoing Support : Offering continuous support to refine and improve the models based on feedback.

Fraunhofer Gesellschaft Zur Forderung Der Angewandten Forschung Ev (Fraunhofer)

**Who Can Benefit:** - MedTech Companies Seeking Certification: For independent validation evidence required by notified bodies and regulatory authorities. - Clinical Decision-Makers: For objective assessment of AI tool performance before clinical adoption. - Quality & Regulatory Teams: For comprehensive documentation aligned with EU AI Act, MDR Annex XIV, and relevant AI standards. - Investors & Partners: For independent third-party performance verification as part of due diligence. **Key Features:** - Independent, unbiased evaluation of AI model performance on clinical data - Comprehensive quality assessment: accuracy, robustness, fairness, explainability, and failure mode analysis from established assessment framework - Alignment with regulatory frameworks (EU AI Act, MDR, FDA AI/ML guidance) and standards (ISO 24028, CLAIM checklist) - Secure on-premises evaluation – no cloud computing; models and data never leave secure infrastructure - Clinical relevance assessment by domain experts **Possible Applications:** - Fairness & Bias Audits: Independent assessment of whether AI models perform equitably across demographic subgroups (age, sex, ethnicity) in clinical and sports populations. - Pre-Submission Validation (MedTech): Independent performance verification of cardiac AI algorithms (e.g., arrhythmia detection) prior to MDR or FDA regulatory submission. - Post-Market Performance Monitoring: Periodic re-validation of deployed AI models to ensure sustained accuracy as patient populations or data acquisition hardware evolve. - Biomechanical AI Benchmarking: Objective validation of movement analysis algorithms against gold-standard motion capture in laboratory and real-world conditions. - Neurological AI Quality Assessment: Stress-testing seizure detection or tremor classification models across diverse patient populations and recording conditions. - Sports & Rehabilitation Algorithm Verification: Validating AI-based injury risk scores, return-to-play classifiers, or performance metrics against clinical expert assessments and ground truth outcomes. - Comparative Algorithm Studies: Head-to-head benchmarking of competing AI solutions on the same clinically curated dataset to support procurement or partnership decisions. **Who We Are:** The **Fraunhofer Insitute for Integrated Circuits (Fraunhofer IIS)** has established the **"Center for Sensor Technology and Digital Medicine" (CEMDIS)** in cooperation with the Universitätsklinikum Erlangen and the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) to enhance modern healthcare through **innovative sensor technology** and **digital solutions**. This center focuses on integrating **innovative medical technologies** such as **wearables** and **robotic systems** to support **medical diagnostics**, **patient monitoring** and **evaluating patient-specific therapies** by providing digital health solutions für real-life healthcare. Located at the Universitätsklinikum Erlangen, it offers unique infrastructures for the **development**, **integration**, and **validation** of novel health technologies, providing companies opportunities for **technological advancements**. For more information, visit the [Fraunhofer IIS website](https://www.iis.fraunhofer.de/de/ff/sse/health/zentrum-sensorik-medizin.html).

Universita Degli Studi Di Pavia (UNIPV)

This service offers the construction and simulation of neurons and micro-circuits for neurological and neuropharmacological applications. The models can be combined to develop and simulate brain digital twins of neuropathologies. Method description: Customization of micro-circuits tailored to specific pathological and/or pharmacological needs. These circuits will be refined and optimized for simulations, with the resulting data integrated into a clinical framework and/or used to model pathological behaviors in a virtual brain twin. Method reference: Local method respecting General Data Protection Regulation (GDPR).

Univerzita Komenskeho V Bratislave (UK BA)

This service provides evaluation of the performance of AI algorithms, whether applied to supervised or unsupervised learning tasks. The evaluation encompasses accuracy, reliability, robustness, and calibration, utilizing standard performance metrics, task-specific metrics, SME-supplied data, and relevant publicly available datasets. Furthermore, the assessment includes an analysis of the AI algorithm's performance across various subpopulations.

Physikalisch-Technische Bundesanstalt (PTB)

The behaviour and trustworthiness of a medical AI system is highly dependent on the data it was trained on. The EU AI Act and MDR therefore place quantitative data quality assessment at the core of development, evaluation, and regulation. Our platform "Metric Hub" ([https://metric.ptb.de/](https://tef.charite.de/r/ro6)) is a dedicated entry point for systematic, fit-for-purpose data quality evaluation in medical AI. Built on the METRIC-Framework, it provides an extensive, openly accessible library of evaluation metrics for both developers and regulatory auditors to evaluate the quality of training and test data. Each metric is presented as a detailed "Metric Card" summarizing the most important information (e.g. definition, value range, applicability, pitfalls, prerequisites) to make it easily implementable and interpretable.

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

Support in setting up a privacy-preserving architecture (Secure Platform for Medical Data Analysis) on premise (local private cloud) when anonymisation/ pseudonymisation cannot be guaranteed due to the dataset or processing characteristics. Supports various privacy preserving techniques (federated learning, split learning, …) and especially for multicentric approaches. Method Description: Support to install a secure architecture for multicentric analysis of medical data with a complete federated model execution environment if needed. Depending of the use case, we can propose: * Design of cloud architecture * Design of federated learning architecture * Aggregator optimisation in federated learning * Enhanced security : to secure the model shared between the partner in the coalition. ** Full Homomorphic Encryption (FHE) : perform operations on encrypted data without having to decrypt it. ** Secure Multi-party Computation (SMPC) : enables several parties to work together to perform calculations on shared data without revealing the underlying information. ** Differential Privacy (DP) : Addition of Guissian noise in data to reinforce the protection of privacy. Method reference: https://www.mdpi.com/1424-8220/22/2/450 https://www.sciencedirect.com/science/article/abs/pii/S0020025518308338 https://www.intechopen.com/chapters/45421 https://ercim-news.ercim.eu/en126/special/inah-the-ethical-secure-platform-for-medical-data-analysis

RISE Research Institutes Of Sweden

Method description: Privacy-preserving and Federated analysis of the solution adapted to the customer Data and use cases. The service is customized according to the SME's needs. Method reference: To be refined depending on the type of task, and in dialog with the customer. The Center for Applied AI at RISE carries out cutting-edge research in AI, connects expertise and applications within RISE, and explores a wide range of innovative applications with industry and the public sector. Applied AI Centre at RISE helps companies and government agencies to see more potential in the technology, use it more wisely and develop it faster. Keywords: Trustworthy AI, Federated learning, Data Science, AI, Deep Learning, Natural language processing, Computer Vision, Machine Learning.

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

Development of PoCs (Proof of Concepts) or demonstrators involving data collection from distributed and heterogeneous data sources (IoT sensors, devices, APIs, etc.) based on lightweight but yet flexible and evolvable middleware architecture leveraging data modeling and data semantics. This service is customised to the customer need and PoCs can cover one, part or all of the following : IoT (wireless) protocols interfacing, gateway integration and programming, implementation of data processing, transformation and transmission to remote platforms. PoCs leverage DMWay toolbox where relevant ( https://asset.cetic.be/en/dmway/) Keywords: IoT, medical devices, wireless communication,firmware, middleware, PoC , prototype, gateway

Karolinska Institutet (KI)

The Swedish node is a collaboration between Karolinska Institutet, SciLifeLab and RISE. Together, we offer world-leading services with our unique collection of core facilities. We can grant services in expert consulting, virtual- and physical testing in the range of in vivo imaging, ex vivo OMICS, pharmaceutical development, simulated healthcare environments, AI-system validation and development, advanced data analysis and other data-driven life science. Prospective data set collection: A full-service collection of new data sets. This service serves to create data sets on-demand according to the needs and requirements of the SME for TEF-project purposes, such as validation. Data sets will primarily comprise in vivo imaging data (PET, CT, MRI, MEG, EEG), ex vivo imaging, and OMICS data within neuro and cancer. Data can be generated by CIR and SciLifeLab infrastructures within our “Physical Testing” services or assembled from publicly available databases and research collaborations. We provide expertise and technical support in the following areas: o Project design& management o Ethical application support o Data collection o Data assembly o Data cleaning, annotation, anonymisation This service can typically directly collaborate with “physical testing services” and technology infrastructures to generate raw data material.

HUMANI SC - Hospital network (HUMANI)

Provision of clinical datasets that are curated, annotated by clinical experts, and available under several standardised formats upon request (OMOP CDM, FHIR, etc.). This service provided by clinical and IT experts, is aimed at delivering high-quality labelled real-world datasets to ensure efficient and high-performing AI solutions trainings.

HUMANI SC - Hospital network (HUMANI)

Provision of structured patient datasets organized in cohorts and mapped to international standards upon request (OMOP CDM or FHIR). This service ensures technical interoperability and facilitates multicenter databases validation by providing ready-to-use data for large-scale clinical evaluation.

RISE Research Institutes Of Sweden

RISE Cyber Range is a secure and realistic training and testing environment. The service also includes controlled penetration testing. Organizations can use the environment to practice and evaluate their ability to prevent, detect, and manage cyberattacks. The service is designed to be understandable even for non-technical decision-makers. It provides a clear picture of an organization’s cybersecurity maturity. RISE Cyber Range uses highly realistic IT and OT environments. These environments are used to identify technical vulnerabilities through penetration testing. They are also used to improve incident response capabilities. The service strengthens collaboration between technical teams, management, and business functions. How can the service help you? The service helps organizations to: - Perform penetration testing - Train staff using realistic cyber incident scenarios - Test technical safeguards and processes in a safe environment - Evaluate organizational readiness and decision-making under pressure - Identify improvement areas before real incidents occur How the service will be delivered Logistics: The Cyber Range environment is provided by RISE and can be delivered on-site, remotely, or as a hybrid solution depending on customer needs. Delivery period: Delivery is planned in close cooperation with the customer and can be conducted as a single engagement or as a series of recurring exercises over time. Duration: The duration of the exercises ranges from a few hours to several days, depending on the scenario and desired level of depth. Customer requirement: The customer is expected to provide high-level information about their organization, relevant threat scenarios, and participants for the exercise. Deliveries: RISE delivers a fully configured cyber range environment, scenario-driven attacks, facilitation, and technical and methodological support throughout the engagement. Output: After completion of the service, the customer receives structured feedback in the form of observations, analysis, and recommendations. Service customization The service can be customized to meet the customer's specific needs, technical environment and level of maturity in cybersecurity. The assessment process begins with a joint meeting where the customer and a technical team from RISE discuss different options and tailor a service that is designed based on the customer's needs.

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

Hospital processes optinisation. CETIC offers its expertise in the field of data management, process simulation, and other innovative expertise that could best optimize the multiple flows related to processes in a hospital, such as the sterilization process. The idea is to build a simulation model of the centralized shared sterilization process based on existing data from several hospitals in order to test and optimize the sizing of the future process. How can the service help you? This service provides a simulation service of sterilization facilities, to evaluate the sizing and robustness of such facilities against the expected flow of trays to sterilize. How the service will be delivered? The service will be delivered by first collecting relevant input data, which includes the medical tool documentation (detailing the description of medical tool sets exported from the sterilization software) and activity data (providing traceability of the passage of sets within the sterilization department). Additionally, data representing the sterilization facility and its environment, such as transportation logistics (e.g., by trucks if applicable), will be gathered. Once the input data is collected, a series of scenarios will be simulated to assess the performance and capacity of the sterilization facility. These simulations will validate investment decisions by providing key performance indicators, including the occupancy rate and overall performance of the facility under various operating conditions. Method reference: Discrete event simulation

Technische Universitaet Muenchen (TUM)

In silico training and optimization: the customer will specify an architecture for software controllers for robotic systems, which will then be deployed in concurrent simulations for e.g., training through reinforcement learning or optimization through genetic algorithms. Performance evaluation of AI-based controllers for robotics based on virtual testing environment: the system provided by the customer, for instance a robotic device using an AI module in closed loop with sensors and actuators for sensorimotor control purposes, will be tested using a simulated environment with sufficient physical realism. Method Description: The Neurorobotics Platform (NRP) is a simulation framework for implementation of highly modular, physically realistic simulations, with a focus on robotics. It also supports the execution of such simulations in the context of an online service on EBRAINS, and can be used in conjunction with standard RL or optimization frameworks. It enables the design, functional evaluation and (at the end of TEF-Health) certification activities for robotic systems, including those with AI modules embedded. Method reference: To be refined depending on the type of task. Either ISO references, scientific literature, work done by WP7, or metrics designed by the TEF-Health consortium.

Friedrich-Alexander-Universitaet Erlangen-Nuernberg (FAU)

# Overview This service delivers an accessible, open-source software solution designed to extract, process, and analyse movement and human walking patterns from wearable sensors. By applying advanced algorithmic workflows to raw motion data gathered from inertial measurement units (IMUs), the service automatically calculates key stride, timing, and distance metrics. The primary objective is to give researchers, clinicians, and health tech innovators a standardized, reproducible way to translate raw sensor readings into valuable physical movement insights without building complex data pipelines from scratch. ## How can the service help you? Developing custom movement-tracking algorithms from raw sensor noise requires deep domain expertise and significant development time. This service bridges that gap by transforming raw, unstructured accelerometer and gyroscope readings into validated, actionable gait metrics. - *Before:* Raw, complex, multi-axis IMU streams that require specialized mathematical filtering and manual spatial-temporal extraction. - *After:* Standardized, high-level gait characteristics ready for clinical interpretation, algorithm validation, or product integration. It helps developers and clinical teams accelerate product validation, compare prototype performance, and lower the technical barriers to integrating movement tracking into digital health tools. ## How will the service be delivered? The service is executed remotely as a software-based data processing workflow. Clients submit anonymized raw IMU data files via secure digital channels. After execution, the processed output is returned alongside structured data reports according to agreed project parameters. # Additional information ## Provider description Operating from the Department of Artificial Intelligence in Biomedical Engineering at Friedrich-Alexander-Universität Erlangen-Nürnberg, we are a service provider node within the TEF-Health consortium. The research group specializes in machine learning, biomedical signal processing, and multimodal sensor synchronization. Our team provides testing infrastructure and scientific support for evaluating medical devices, wearables, and contactless sensing systems. ## Technical details The framework is built around **gaitmap**, an open-source Python library tailored for spatial-temporal gait parameter extraction. - *Input Data Requirements:* Raw tri-axial accelerometer and gyroscope data collected from foot, ankle, or lower-body IMUs. Data must be anonymized prior to transfer. - *Core Tasks & Processing:* Signal pre-filtering, stance/stride phase segmentation, step event detection, and spatial parameters calculation (stride length, velocity). - *Outputs:* Structured data exports (e.g., JSON/CSV) detailing gait cadence, stride time, swing phase duration, step length, and walking speed. - *Reference & Code Base:* [https://github.com/mad-lab-fau/gaitmap](https://github.com/mad-lab-fau/gaitmap) ## Service customization While the open-source pipeline offers standard default parameters for adult gait analysis, the pipeline can be customized for specific target populations (e.g., geriatric care, neurological conditions, high-performance sports). SMEs and startups can choose specific processing modules (e.g., requesting only temporal parameters or custom data visualization) rather than executing the full end-to-end pipeline.

Friedrich-Alexander-Universitaet Erlangen-Nuernberg (FAU)

# Overview This service provides an open-source software framework designed to standardize and simplify the evaluation, comparison, and integration of machine learning models alongside traditional algorithms. When working with complex or multi-modal biomedical sensor data, standard evaluation tools often struggle to handle non-tabular data structures and custom processing steps. By utilizing an object-oriented paradigm, the framework enables teams to build modular dataset interfaces and structured execution pipelines. The primary objective is to streamline validation workflows, prevent implementation errors during nested cross-validation or parameter optimization, and provide a unified benchmark platform for comparing legacy algorithms with modern data-driven approaches. ## How can the service help you? Evaluating non-standard machine learning pipelines and comparing them with conventional heuristics often leads to complex, custom codebases prone to data leakage and validation errors. This service establishes an object-oriented architecture to structure your datasets and algorithm steps cleanly. - *Before:* Fragile, unstandardized script setups where comparing machine learning workflows against traditional algorithmic baselines requires manual, error-prone evaluation logic. - *After:* A robust, modular framework with unified interfaces for parameter optimization, cross-validation, and standardized performance comparison. It allows development teams to accelerate algorithm benchmarking, enforce reproducible evaluation standards, and maintain clean abstractions across heterogeneous code bases. ## How will the service be delivered? The service is delivered through software access, technical documentation, and collaborative setup support. Clients provide details regarding their multi-modal data structure and algorithm requirements. Our team assists in configuring the software architecture, defining custom dataset classes, and setting up standardized validation pipelines. # Additional information ## Provider description Operating from the Department of Artificial Intelligence in Biomedical Engineering at Friedrich-Alexander-Universität Erlangen-Nürnberg, we are a service provider node within the TEF-Health consortium. The research group specializes in machine learning, biomedical signal processing, and multimodal sensor synchronization. Our team provides testing infrastructure and scientific support for evaluating medical devices, wearables, and contactless sensing systems. ## Technical details The service relies on **tpcp** (Tiny Pipelines for Complex Problems), an open-source Python framework tailored for complex algorithm evaluation. - *Input Data Requirements:* Complex, multi-modal, or time-series datasets requiring custom data structures and custom splitting/indexing logic. - *Core Tasks & Processing:* Implementation of object-oriented dataset interfaces, creation of modular algorithm pipelines, parameter optimization, nested cross-validation, and performance benchmarking across diverse model types. - *Outputs:* A structured software framework containing object-oriented data loaders, standard pipeline interfaces, and reproducible evaluation tools tailored to client datasets. - *Reference & Code Base:* [https://github.com/mad-lab-fau/tpcp](https://github.com/mad-lab-fau/tpcp) ## Service customization The engagement can be tailored to the client's current software stack. SMEs and startups can choose to adopt individual components (e.g., utilizing specific dataset helpers) or engage in building full parameter optimization and multi-model benchmark suites customized for their proprietary algorithms.

Health Data Hub (Plateforme des Données de Santé)

This services include the following : Project launch : assistance in framing the project, particularly in identifying and implementing regulatory procedures ; establishment of contacts with relevant stakeholders ; support in collaborating with bodies in charge of health data, particularly with regards to contracting and meeting technical prerequisites for using such data ; Support towards institutional actors, particularly when several data sources are involved and need to be combined or synchronized ; Data handling : support for data custodians in extracting, structuring, qualifying, documenting, standardizing and/or transferring data involved as part of the project ; Support to research implementation : provision of technological capabilities (dedicated project spaces with computing resources for 3 years, state of the art data management, analysis and visualization tools, alongside user onboarding and ongoing technical support), expert support (medical, legal, data scientist, engineer, etc.), training and networking ; Results and project valorization : communication events, networking within the ecosystem, etc. Catalogue database fee : manages the fees related to accessing and utilizing electronic health data for secondary use. The coverage is partial and depends on the chosen database, including standard access fees to database holders, potential compensation for a portion of the data collection costs, and a specific fee covering human and technical resources expended in enriching the electronic health data.

Health Data Hub (Plateforme des Données de Santé)

This services include the following : Project launch : assistance in framing the project, particularly in identifying and implementing regulatory procedures ; establishment of contacts with relevant stakeholders ; support in collaborating with bodies in charge of health data, particularly with regards to contracting and meeting technical prerequisites for using such data ; Support towards institutional actors, particularly when several data sources are involved and need to be combined or synchronized ; Data handling : support for data custodians in extracting, structuring, qualifying, documenting, standardizing and/or transferring data involved as part of the project ; Support to research implementation : provision of technological capabilities (dedicated project spaces with computing resources for 3 years, state of the art data management, analysis and visualization tools, alongside user onboarding and ongoing technical support), expert support (medical, legal, data scientist, engineer, etc.), training and networking ; Results and project valorization : communication events, networking within the ecosystem, etc. Catalogue database fee : manages the fees related to accessing and utilizing electronic health data for secondary use. The coverage is partial and depends on the chosen database, including standard access fees to database holders, potential compensation for a portion of the data collection costs, and a specific fee covering human and technical resources expended in enriching the electronic health data.

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

Help to select and to use advanced deep learning algorithms to create custom and trustworthy AI models for medical image analysis (e.g. classification, segmentation, feature extraction, anomaly localization). Depending on the modality to be processed, we propose advanced deep learning architectures (CNN, LSTM, ViT, etc.) by selecting the best approach for the task to be performed (classification, localisation, segmentation, etc.). Keywords: Medical Image Analysis; Deep Learning Algorithms; Binary Classification; Anomaly Localization; AI Model Development; Real-World Images; AI Testing and Validation; Collaboration with Medical Staff; AI Expertise; Custom Pretrained Pipelines; Clinical Imaging; Image Segmentation; Image Feature Extraction; Trustworthy AI

Politecnico Di Milano (POLIMI)

This service assesses algorithm performance, development progress, and adherence to AI Act regulations for implementation in clinical practice. The AI validation includes several aspects, aligned with TEF guidelines, and EU regulation: 1) Bias evaluation 2) Accuracy evaluation 3) Trustworthiness evaluation 4) Uncertainty evaluation. The validation will be offered on three levels, based on the source of the validation dataset: 1) End-user provides the data for validation (input features) 2) End-user provides data for validation at the start from raw data of the initial dataset (data curation and feature extraction on the validation set will be extracted by the Lab within the process of validation) 3) The POLIMI Lab provides a tailored external validation set acquired from other clinical centers. Keywords: Validation; Bias; Accuracy, Trustworthiness, Uncertainty, AI algorighm AI validation

Fraunhofer Gesellschaft Zur Forderung Der Angewandten Forschung Ev (Fraunhofer)

This module assesses whether the uncertainty estimates provided by an AI model are well-calibrated, meaningful, and suitable for supporting safe decision-making in clinical or high-stakes contexts. The customer-provided model's uncertainty outputs (e.g., confidence scores, predictive distributions, epistemic/aleatoric uncertainty estimates) are evaluated for calibration, sharpness, and reliability. Evaluation can be performed with or without ground truth labels, depending on availability: • With ground truths: Direct assessment of uncertainty quality can be conducted by evaluating how well uncertainty scores align with truly uncertain outcomes. This includes analysis on real out-of-distribution (OOD) samples and ambiguous cases (e.g., annotator disagreement), enabling a precise evaluation of whether high uncertainty corresponds to genuinely difficult or uncertain predictions. • Without ground truths: Proxy tasks can be employed to assess uncertainty quantification capabilities, including synthetic out-of-distribution sample generation, Expected Calibration Error (ECE), Prediction Rejection Ratio (PRR) analysis, and correctness prediction assessment. While less direct, these approaches still provide meaningful insights into model uncertainty quality. Evaluation Outcomes: Positive Indicators • Well-calibrated confidence scores (predicted probabilities reflect true outcome frequencies) • Appropriate uncertainty increases for ambiguous, atypical, or out-of-distribution inputs • Meaningful separation between epistemic and aleatoric uncertainty (if applicable) Negative Indicators • Overconfident predictions on uncertain or out-of-distribution cases • Poorly calibrated probability estimates • Uncertainty scores that do not correlate with prediction errors

Karolinska Institutet (KI)

Overview Usability Testing of Digital Systems is a tailored service that recruits representative end users to test your product in real world conditions and delivers a comprehensive usability report with actionable insights on user experience, performance, and relevance to help you refine and improve your product before wider launch. Our beta testers are carefully selected to encompass a broad and diverse range of fields representative of your software solution end-users. Your product will be tested in real-world conditions to provide you with a panel of key parameters essential to either fine-tune, improve or adjust various features of your solution. Following the testing, we will provide you a summative evaluation encompassing the following assessments and feedback for: Usability: user experience (UX) and overall satisfaction while using your solution Performance: speed, efficiency and accuracy for the different operations generated Relevancy and limitations: highlight which fields benefit most of the use of your software and the potential application areas lacking information or irrelevant for your solution. This service allows you to highlight the strength and weakness of the current state of your product and thus to provide you the opportunity to address them according to the direct feedback from your end-users, and to improve your product before a broader public launch. How the service will be delivered? Step 1: Assessment of your product, functionality and purpose Step 2: Identification of the scope of your end-users Step 3: Participants recruitment (from healthcare professionals and/or research community, according to the need) Step 4: Design of the protocol to conduct usability test Step 5: Data collection and summative evaluation report Provider description The Swedish TEF-Health node is a collaboration between Karolinska Institutet, SciLifeLab and RISE, and is led by Karolinska Institutet. Together, we offer world-leading services with our unique collection of core facilities. We can grant services in expert consulting, virtual- and physical testing in the range of in vivo imaging, ex vivo OMICS, pharmaceutical development, simulated healthcare environments, AI-system validation and development, advanced data analysis and other data-driven life science.

Fraunhofer Gesellschaft Zur Forderung Der Angewandten Forschung Ev (Fraunhofer)

This module evaluates whether AI-generated explanations and outputs are understandable, usable, and actionable for the intended human stakeholders (e.g., clinicians, radiologists, patients). The module examines usability, perceived fairness, responsibility & control (human vs. AI), cognitive load, and the emotional and ethical impact of AI in the clinical context. Method Description The evaluation employs established usability and human-centered AI methods, which may include: • Heuristic evaluation: Expert review of AI interfaces against usability, user experience & design principles • Cognitive walkthrough: Task-based analysis of how users would interpret and act on AI outputs (e.g. at critical decision points, including situations where human and AI recommendations diverge) • User studies (optional, scope-dependent): Structured interviews, think-aloud protocols, or questionnaire-based assessment with representative end users to capture variables like understanding, trust, and perceived fairness. • Quantitative measurements: Trust and workload assessments to quantify trust calibration and cognitive load. • Scenario-based evaluation of clinical workflows for edge cases and high-risk situations, to assess human oversight and potential over-/under-reliance on AI • Workshops (optional, scope-dependent): Value- and risk-focused workshop with different stakeholders to investigate risks, and requirements for human oversight. The specific methods applied are tailored to the customer's use case, user groups, and available resources. Evaluation Outcomes Positive Indicators • Explanation quality: Explanations are comprehensible to the intended user group without requiring AI expertise. Users can appropriately calibrate trust based on provided explanations. Explanation format and complexity match clinical workflow requirements • Roles and responsibilities between humans and the AI system are clearly understood and accepted by users • Users perceive the systems behaviour and explanations as fair and aligned with clinical values and ethical standards Negative Indicators • Explanation quality: Explanations are misleading, overly technical, or ambiguous. Users misinterpret explanations or develop inappropriate trust/distrust. Explanation presentation disrupts clinical workflow or decision-making • Unclear responsibility between human and AI leads to uncertainty in high-stakes decisions or error handling • AI is perceived as biased or unfair toward certain patient groups • Increased stress, uncertainty, or ethical issues among clinicians or patients due to the AI system This service is delivered by Fraunhofer HHI as a TEF-Health partner. Fraunhofer HHI provides: • Dedicated compute infrastructure • In-house proprietary XAI and evaluation software • Expert scientific and technical support across all evaluation modules

Centro Hospitalar De Sao Joao Epe (CHSJ)

The service offers SMEs access to hospital workflows and infrastructure within a controlled, virtual environment. By simulatin a hospital setting, it assesses de the compatibility, interoperability and reliability of systems before deployment in real-world scenarios, reducing risk, saves resources and streamlines the integration proccess by identifying and resolving issues early in the development lifecycle.

Centre D'Excellence En Technologies De L'Information Et De La Communication (CETIC)

To enhance the security of software-based products, ensure compliance, and mitigate risks, CETIC offers expertise in scanning your system with automated tools to identify known vulnerabilities and potential weak points that attackers could exploit. This method involves the use of specialized software to identify security flaws, such as obsolete software versions, incorrect configurations or unsecured open ports. Once these vulnerabilities have been identified, an action plan can be drawn up, including security measures to reduce the risks. We use industry standards such as OWASP Top 10 to configure the scanning process. How can the service help you? We will provide complete scan reports as well as recommendations to lower the residual risk level and attack surface of your system. Together with a risk analysis, they can be used as a complete set of evidences towards authorities and customers. How the service will be delivered? Our team will carry out comprehensive vulnerability analysis for your software product, and help you define the most appropriate action plan, taking into account your specific requirements and context. We scan your system using automated tools to detect known vulnerabilities or weak points that could be exploited by attackers. This method involves the use of specialized software to identify security flaws, such as obsolete software versions, incorrect configurations or unsecured open ports. Once these vulnerabilities have been identified, we define with you actionable recommendations including security measures to reduce the risks. Optionally, we can complement with an analysis of the source code by a tool that will scan the whole codebase searching for security violations. This will further improve the cyber-resilience of your product. Optionally, we can perform the security risk analysis of your product. Service deployment: The service is usually "deployed" by simply having remote access to your product, so as to execute the vulnerability scanning. For the optional source code analysis, we need access to the code base. Resources provided to client: Cybersecurity tests report Recommendations Method reference: OWASP Top 10