Work Packages
BRAINTWIN is organized into seven work packages (WPs) that break down the project into manageable tasks with clear objectives and deliverables. Each WP is led by a consortium partner with contributions from multiple institutions.
Data Integration & FAIR Platform
Duration: Months 0-48
Lead: ICM (Paris Brain Institute) with HCL, CentraleSupelec, Ecole Polytechnique
Objective: Establish BRAINTWIN's data foundation by aggregating, harmonizing, and deploying a secure FAIR-compliant platform for multimodal brain tumor and SVD datasets.
Key Tasks:
- Task 1.1 - Multi-source Data Collection: Compile ~3,000 brain tumor cases and ~1,000 SVD cases with MRI, histopathology, molecular data, and clinical records
- Task 1.2 - Data Harmonization: Standardize using BIDS for MRI and OMOP for clinical data; perform preprocessing, anonymization, and validation
- Task 1.3 - FAIR Platform Development: Build secure federated platform with role-based access, GDPR compliance, and ethical governance
Multimodal AI Methods Development
Duration: Months 0-48
Lead: CentraleSupelec (Prof. Vakalopoulou, Prof. Christodoulidis) with Polytechnique, INSA Lyon, ICM
Objective: Develop novel algorithms to extract features from each data modality and fuse them into an integrated model for clinically useful predictions.
Key Tasks:
- Task 2.1 - Multimodal Fusion Model: Develop architecture merging features across modalities using early/late fusion, cross-attention transformers, and dynamic fusion
- Task 2.2 - Outcome Prediction: Train prediction heads for tumor subtype classification, survival time, and SVD risk with uncertainty estimation
- Task 2.3 - Explainability: Implement XAI methods (Grad-CAM, attention heatmaps) for interpretable predictions
Federated Learning & Privacy
Duration: Months 6-48
Lead: Ecole Polytechnique (Prof. Dieuleveut) with Dr. Even, Prof. Bellet
Objective: Adapt federated learning methodology to multiple modalities and deploy a privacy-preserving FL framework across sites.
Key Tasks:
- Task 3.1 - FL Framework Setup: Configure federated platform (FedML/Flower) with secure server-client communication
- Task 3.2 - Algorithmic Enhancements: Explore FedProx, FedDyn, meta-learning for data heterogeneity
- Task 3.3 - Privacy Preservation: Implement differential privacy and robust aggregation techniques
- Task 3.4 - Federated Training: Train BRAINTWIN model on distributed datasets and compare with centralized models
Clinical Validation: Neuro-Oncology
Duration: Months 12-48
Lead: ICM & Hospices Civils de Lyon (Dr. Alentorn, Prof. Ducray)
Objective: Rigorously evaluate the BRAINTWIN digital twin in brain tumor patients and demonstrate clinical utility.
Key Tasks:
- Task 4.1 - Retrospective Validation: Test model on retrospective cohorts, comparing against standard prognostic scores
- Task 4.2 - Prospective Pilot: Deploy system in clinical workflow at partner hospitals
- Task 4.3 - Biomarker Discovery: Uncover novel radiogenomic associations and prognostic biomarkers
- Task 4.4 - Clinical Metrics: Evaluate health-economic impact and QALY gains
Clinical Validation: SVD
Duration: Months 12-48
Lead: ICM (Prof. Debette) with VBHI, Dr. Joliot
Objective: Test and validate the digital twin approach for cerebral small vessel diseases and neurodegenerative disorders.
Key Tasks:
- Task 5.1 - Cohort Analysis: Format cohort data and adapt model for stroke recurrence, cognitive decline prediction
- Task 5.2 - Outcome Validation: Evaluate predictive performance vs Framingham Stroke Risk Score
- Task 5.3 - International Cohorts: Test generalizability using ISGC/CHARGE international datasets
- Task 5.4 - Prevention Strategies: Use digital twin to simulate interventions and identify personalized prevention
Dissemination & Exploitation
Lead: ICM (Dr. Alentorn) with all partners
- Publications and conference presentations
- Open-source software toolkit release
- Stakeholder engagement and workshops
- Exploitation and sustainability planning
Project Management
Lead: ICM (Dr. Alentorn, Coordinator)
- Steering Committee governance
- Reporting and documentation
- Risk management
- External Scientific Advisory Board