0604

View Abstract / View Presentation

Disability Prediction in Multiple Sclerosis using Ensemble of Machine Learning Models and DTI Brain Connectivity

Berardino Barile¹, Aldo Marzullo², Claudio Stamile³, Françoise Durand-Dubief⁴, and Dominique Sappey-Marinier^1,5
¹CREATIS (UMR 5220 CNRS & U1206 INSERM), Université Claude Bernard Lyon 1, Villeurbanne, France, ²Department of Mathematics and Computer Science, University of Calabria, Rende, Italy, ³R&D Department, CGnal, Milan, Italy, ⁴Hôpital Neurologique, Hospices Civils de Lyon, Bron, France, ⁵MRI, CERMEP - Imagerie du Vivant, Bron, France

The proposed Stacking Ensemble scheme provided excellent prediction performance in predicting MS disability, using connectome data and fiber-bundles data. The counterfactual model highlighted WM links usually associated with disability increasing the accountability of the method.

Figure 1: Pipeline of the proposed Ensemble and Interpretability models for EDSS prediction and visualization of the brain networks responsible for the prediction.

Figure 2: Comparison of measured (red) and estimated (blue) EDSS score in MS patients.