Good Features for Reliable Registration in Multi-Atlas Segmentation

Abstract

This work presents a method for multi-organ segmentation in whole-body CT images based on a multi-atlas approach. A robust and efficient feature-based registration technique is developed which uses sparse organ specific features that are learnt based on their ability to register different organ types accurately. The best fitted feature points are used in RANSAC to estimate an affine transformation, followed by a thin plate spline refinement. This yields an accurate and reliable nonrigid transformation for each organ, which is independent of initialization and hence does not suffer from the local minima problem. Further, this is accomplished at a fraction of the time required by intensity-based methods. The technique is embedded into a standard multi-atlas framework using label transfer and fusion, followed by a random forest classifier which produces the data term for the final graph cut segmentation. For a majority of the classes our approach outperforms the competitors at the VISCERAL Anatomy Grand Challenge on segmentation at ISBI 2015.