An experimental study on combining the auto-context model with corrective learning for canine LEG muscle segmentation

Abstract

Corrective learning is a technique that applies classification methods for automatically detecting and correcting systematic segmentation errors produced by existing segmentation methods with respect to some gold standard (manual) segmentation. To allow corrective learning more effectively correct errors that require non-local contextual information to capture, we extend the corrective learning technique by combining it with auto-context learning and conduct experimental study to verify its effectiveness. In our experiment, we take multi-atlas joint label fusion as the host segmentation method, for which we apply our corrective learning technique to improve, and apply it on a canine leg muscle segmentation application. We show that the auto-context enhanced corrective learning produces prominent improvement over the original corrective learning method.