Automatic Multiple Sclerosis detection based on integrated square estimation

This paper presents a fully automatic method for segmentation of multiple sclerosis (MS) lesions from multiple sequence MR (T2-weighted and FLAIR) images. Our method treats MS lesions as outliers to the normal brain tissue distribution, and the separation is achieved by minimizing a statistically robust L₂E measure, which is defined as the squared difference between the true density and the assumed Gaussian mixture. Pre- and post-processing procedures including intensity normalization and false positive pruning are designed to remove various signal artifacts. Our method is fully automatic and doesn´t require any training, atlas or thresholding steps. The results of our method are compared with lesion delineations by human experts, and a high classification accuracy is demonstrated on 16 datasets containing small to moderate lesion loads.