Probabilistic Spatial-Temporal Segmentation of Multiple Sclerosis Lesions

Abstract

In this paper we describe the application of a novel statistical video-modeling scheme to sequences of multiple sclerosis (MS) images taken over time. The analysis of the image-sequence input as a single entity, as opposed to a sequence of separate frames, is a unique feature of the proposed framework. Coherent space-time regions in a four-dimensional feature space (intensity, position (x, y), and time) and corresponding coherent segments in the video content are extracted by unsupervised clustering via Gaussian mixture modeling (GMM). The Expectation-Maximization (EM) algorithm is used to determine the parameters of the model according to the maximum likelihood principle. MS lesions are automatically detected, segmented and tracked in time by context-based classification mechanisms. Qualitative and quantitative results of the proposed methodology are shown for a sequence of 24 T2-weighted MR images, which was acquired from a relapsing-remitting MS patient over a period of approximately a year. The validation of the framework was performed by a comparison to an expert radiologist’s manual delineation.