Assessing Real-World Performance of AI-Based Neuroimaging Diagnosis Tools T

Title: Assessing Real-World Performance of AI-Based Neuroimaging Diagnosis Tools Through Hospital Data Warehouses

Abstract: The integration of artificial intelligence and neuroimaging opens new perspectives for diagnosing neurological diseases. However, most machine learning approaches have been validated on research datasets, and their generalisation to clinical routine remains unclear. Hospital data warehouses offer considerable volumes of real-world data, enabling evaluation in routine care settings. We are currently exploiting data from the Greater Paris hospitals (AP-HP, comprising 39 hospitals and millions of patients) to develop AI tools for biomarker identification and disease prediction. This presentation will address the challenges we face related to imaging data quality and heterogeneity, and highlight the solutions we propose.

Speaker Bio: Ninon Burgos is a CNRS research director at the Paris Brain Institute in France, co-head of the ARAMIS Lab, and PR[AI]RIE fellow. Her research focuses on medical image processing, anomaly detection, and translating machine learning approaches into clinical practice. She develops computer-aided diagnosis tools and open-source software for neuroscience applications, enhancing reliability and scalability in clinical neuroscience research.