A Fully Automated, Hierarchical Classification Method for Detecting White Matter Lesions in Multiple Sclerosis

Keywords

although amount applied approaches arrays artificially assigned attempts automated behavioral blue border bottom capable classification classifier classifiers cleaner clear clearly clinical cluster clustered clusters column count counts dataset demonstrating detecting development discovery discrimination dividing done enough errors even example exterior false feature features fewer fluid fully future goal good great grossly hidden hierarchical highlighting important improve influence initial inputs inside intensities intensity intervention investigate just layer lesion lesions linearly made maintaining manual manually many mark measured measures median minor much negative negatives network neural node normalized note novel obviously offers optimize output outputs overestimated overlap oxford patches patients percentiles performance plus positive positives potential preprocessing probabilities probability problem propose proposed quantitative random reduces reducing reduction reductions registered representative restricted sample scale scheme sclerosis segmentation segmentations segmented selected separately sets slice statistics studies subjects subset substantial task terms threshold thresholded thresholding thresholds tissue training transformed true types units user white wise yellow