Robust Image Segmentation through Outlier Handling Application to Automatic Plaque Burden Assessment
Merrifield R, Lekadir K, Yang G, Pennell D, Keenan N
This abstract presents a robust automatic segmentation method that can be used for automated plaque burden assessment using CMR. Manual delineation can be problematic as it is time consuming and subject to significant operator bias. On the other hand, existing automatic techniques suffer from lack of accuracy when some landmarks lie on incorrect boundary positions due to noise and artefacts. The proposed method is able to detect and correct these outliers using an invariant shape representation. The validation on real data demonstrates high accuracy and robustness. The result shows a potential clinical value for longitudinal studies of plaque burden.