Calibrating Slicer's EM Segmenter

The Slicer module EMAtlasBrainClassifier EMAtlasBrainClassifier robustly partitions MRIs into gray and white matter, and cortical spinal fluid. Many labs, however, are interested in further dividing these structures or identify pathologies, such as lesions. Also, the module currently requires T1 and T2 weighted images. In this tutorial, we explain, how to customize the module to these segmentation problems.

We provide an example of an image sequence, which cannot be segmented with the EMAtlasBrainClassifier module. We go over the necessary steps to adjust the module to this new sequence via the Slicer module EMSegment . EMSegmenter is a more complex version of EMAtlasBrainClassifier. It allows the user to change the image sequences, the number of structures to be segmented, as well as the characterizes of the structures. We end the discussion by showing how to transfer the results of EMSegment to the EMAtlasBrainClassifier module.

We also invited Istvan Csapo (Center for Neurological Imaging , BWH), who will discuss his experience in customizing the module towards healthy aging. Istvan has started a document describing the module, which we hope will be completed by other users.