A Novel 3D Physics Informed Neural Network for Medical Imaging

Información del proyecto

NosoScan

Identificador del acuerdo de subvención: 101310167

DOI

10.3030/101310167

Fecha de la firma de la CE 17 Julio 2026

Fecha de inicio 1 Agosto 2026

Fecha de finalización 31 Julio 2028

Financiado con arreglo a

• Widening participation and spreading excellence

Coste total

Sin datos

Aportación de la UE

€ 499 418,50

Coordinado por

CORTISCOPE S.R.L.
Romania

Objetivo

NosoScan is a next-generation, physics-informed MRI segmentation engine designed to deliver fast, reproducible and regulator-grade tumour volumetrics across heterogeneous clinical environments. Today, neuro-oncology and clinical trials depend on MRI data coming from multiple hospitals, scanners and protocols, creating major variability in tumour measurements, delays in interim analyses, and high re-read costs. Existing AI solutions rely on GPU-heavy black-box models that require isotropic resampling, reduce anatomical fidelity, and lack the auditability required for clinical and regulatory workflows. NosoScan introduces a mm-native, physics-in-the-loop architecture that preserves true MRI geometry and ensures high performance even on anisotropic scans. Unlike conventional deep-learning systems, it runs entirely on standard CPUs while maintaining accuracy (Dice ≥0.88) enabling scalable, on-premise deployment in hospitals, CROs and mobile imaging settings without cloud dependency or specialised hardware. The system produces complete volumetric outputs, tumour sub-region maps, uncertainty visualisations, and structured DICOM-SR/PROV provenance. During this project, Cortiscope will advance NosoScan from TRL4 to TRL6 by maturing its physics-guided core, validating performance on retrospective neuro-oncology MRI cases, and delivering an audit-ready MVP. A secure, EU-hosted SaaS/API layer will support interoperability, regulatory alignment and scalable integration into clinical-trial workflows. NosoScan directly addresses unmet needs in clinical research and hospital imaging: reducing cost per scan by up to 85%, enabling faster interim trial readouts, ensuring reproducible endpoints, and supporting data-sovereign deployment across Europe. The project positions Cortiscope for EIC Accelerator funding, CE-marking preparation, and broader clinical and OEM adoption, establishing a European benchmark for trustworthy, transparent and operationally accessible medical-imaging AI.

Ámbito científico (EuroSciVoc)

• ciencias naturales › informática y ciencias de la información › inteligencia artificial › inteligencia computacional

Palabras clave

• Medical Imaging AI
• Brain Tumour Segmentation
• CPU-First Inference
• Physics-Informed Neural Networks
• MRI Anisotropy
• Clinical Trial Imaging
• Neuro-Oncology.

Programa(s)

• HORIZON.4.1 - Widening participation and spreading excellence PROGRAMA PRINCIPAL
  

Tema(s)

• HORIZON-WIDERA-2025-02-ACCESS-01 - EIC pre-accelerator - Widening
  

Régimen de financiación

HORIZON-CSA - HORIZON Coordination and Support Actions

Convocatoria de propuestas

HORIZON-WIDERA-2025-02

Coordinador

CORTISCOPE S.R.L.

• Aportación neta de la UEn: € 499 418,50
• Dirección: Str.Faurei, nr.5, Bl. C9, entrance A, ap. 3 ; 013181 Bucharest ; Rumanía
• Pyme Sí:
  Región
  Macroregiunea Trei › Bucureşti-Ilfov › Bucureşti
• Región: Macroregiunea Trei › Bucureşti-Ilfov › Bucureşti
• Tipo de actividad: Private for-profit entities (excluding Higher or Secondary Education Establishments)
• Coste total: Sin datos