Decoding post-stroke motor function from structural brain imaging.

摘要: Clinical research based on neuroimaging data has benefited from machine learning methods, which have the ability to provide individualized predictions and account for interaction among units of information in brain. Application structural imaging investigate diseases that involve brain injury presents an additional challenge, especially conditions like stroke, due high variability across patients regarding characteristics lesions. Extracting anatomical images a way translates damage into features be used as input algorithms is still open question. One most common approaches capture regional obtain lesion load per region (i.e. proportion voxels structures are considered damaged). However, no systematic evaluation yet been performed compare this approach with using patterns considering each voxel single feature). In paper we compared both applying Gaussian Process Regression decode motor scores 50 chronic stroke solely derived MRI. For different ways delimit areas: regions interest atlas, corticospinal tract, mask obtained fMRI analysis task healthy controls selected lesion-symptom mapping. Our showed extracting through represent probability produced better results than quantifying region. particular, areas compared, best performance was combination range cortical subcortical well tract. These will inform appropriate methodology predicting long term outcomes early post-stroke imaging.