N°250 – A connectome-based approach for lesion-symptom mapping in stroke patients with motor deficits
Résumé
Objectives
Lesion-symptom mapping (LSM) is widely used to explore the association between focal structural damage and functional deficits in stroke patients. In this study, we developed an LSM approach based on changes in the topological properties of lesion-derived brain networks in stroke patients with motor deficits.
Content
Structural MR (3D T1-weighted) from 340 stroke patients (63.9 ± 10.5 years) were included in this work from the multicenter GRECogVASC study conducted at Amiens University Hospital. After a neuropsychological examination, a lesion mask was generated for each patient in the MNI space. For connectivity analysis, 1,000,000 streamlines were generated using the deterministic tractography method using diffusion-weighted imaging data from 400 healthy controls (62.87 ± 13.47 years) included in the CamCAN repository. Using a high-resolution atlas including 1133 regions, a connectivity matrix was then constructed for each healthy subject based on the number of streamlines connecting parcels (nodes). An average lesion-derived connectivity matrix was computed over healthy subjects for each patient by removing white matter pathways passing through the patient’s lesion. The associations between changes in each node’s degree and left/right motor deficits were then computed across patients using the multivariate support vector regression analysis and permutation testing (p < 0.005, 1000 permutations). Using the proposed method, cortical nodes showing significant associations with motor deficits were mainly located within the precentral areas for both the left and right motor deficits. The connectome-based lesion-symptom mapping could identify brain regions associated with motor deficits with significantly reduced false positives in comparison with the conventional univariate regression analysis.