Multi-contrast Patchmatch Algorithm for Multiple Sclerosis Lesion Detection

Due to their abnormal appearance, Multiple Sclerosis lesions can influence the results of various image analysis techniques such as segmentation and registration. As the multi-modal characteristic intensity of the Multiple Sclerosis lesions is different that of non-pathological tissues, a local multi-modal intensity similarity can be used to classify and segment lesions. In this work, lesions are segmented using a fast patch matching approach, namely the optimised PatchMatch label fusion algorithm. The optimised PatchMatch label fusion algorithm is here extended to multimodal data, enabling an accurate Multiple Sclerosis lesion segmentation.