Segmentation of inflamed synovia in multi-modal 3D MRI

We address the difficult problem of segmenting the inflamed synovial tissue in multi-modal 3D MR sequences of the wrist. The complex morphology of the structures to segment, the dimensionality of the images, and the multiple modalities all contributes to the difficulty of the problem. We propose a solution based on a voxel classifier built on relatively few features selected from a large pool computed from all available modalities. The method appears to behave well and matches the performance of human experts. Furthermore, it is a general segmentation scheme which can be easily adapted to other clinical problems.