Statistical and deformable model approaches to the segmentation of MR imagery and volume estimation of stroke lesions

Abstract

We propose two 3D methods to segment magnetic resonance imagery (MRI) of ischemic stroke patients into lesion and background, and hence to estimate lesion volumes. The first is a hierarchical, regularized method based on classical statistics that produces a rigorous confidence interval for lesion volume. This approach requires a limited amount of user interaction to initialize, but this step can be time-consuming. The second method integrates the first into the deformable models framework. This hybrid approach combines intensity-based information provided by the statistical method and shape-based information given by the deformable model. It also requires less initialization than the statistical method. Both procedures have been tested on real MR data, with volume estimates within 20% of those derived from doctors’ hand segmentations. According to the physicians with whom we are working, these results are clinically useful to evaluate stroke therapies.