MEDIFUSE-AI
Experimental development of a transparent multimodal AI platform for clinical decision support in non-small cell lung cancer.
Experimental Development
MEDIFUSE-AI
AI-driven multimodal platform for integrating and interpreting heterogeneous clinical data in non-small cell lung cancer
Experimental development of a transparent multimodal AI platform for clinical decision support in non-small cell lung cancer.
Overview
MEDIFUSE-AI is a 24-month experimental development project co-financed by the European Union through STEP Lazio (FESR Lazio 2021-2027). It brings together Eustema, UCBM-ArCo, and the Fondazione Policlinico Universitario Campus Bio-Medico to develop a multimodal AI platform for the clinical pathway of non-small cell lung cancer.
In current oncology workflows, clinicians must manually combine radiological images, reports, laboratory data, clinical records, and genomic information. MEDIFUSE-AI aims to fuse these sources into transparent, evidence-based decision support and validate the resulting prototype in a real clinical setting at Policlinico UCBM.
Expected results include a secure clinical data infrastructure, explainable multimodal AI modules, a validated prototype for clinical decision support, and a roadmap towards reliable and certifiable deployment, while considering GDPR, the EU AI Act, and the requirements of medical-device development.
Project Objectives
- Build an enterprise-grade clinical data infrastructure for heterogeneous oncology data.
- Develop transparent multimodal AI modules that combine imaging, text, clinical, laboratory, and genomic information.
- Validate the integrated prototype with real clinical cases at Policlinico UCBM.
- Establish a roadmap towards reliable and certifiable clinical deployment.
Implementation Phases
- Clinical needs and data infrastructure
- Trustworthy multimodal AI and explainable outputs
- Platform integration
- Clinical validation and certification roadmap
Research Directions
- Multimodal integration of imaging, clinical records, laboratory data, genomic information, and text.
- Trustworthy and explainable AI for clinical decision support in NSCLC.
- Secure data infrastructure, NLP, multimodal prediction, and generative methods for incomplete clinical data.
- Clinical validation and a roadmap towards medical-device certification.