Towards an Accurate and Generalizable Multiple Sclerosis Lesion Segmentation Model Using Self-Ensembled Lesion Fusion | IEEE Conference Publication | IEEE Xplore