Simultaneous segmentation and tumor detection in MRI cervical cancer radiation therapy with Hierarchical Adaptive Local Affine Registration

Abstract

External Beam Radiation therapy (EBRT) for the cancer treatment enables accurate placement of radiation dose on cancerous region. The presence and regression of tumors may violate registration constraints and cause registration errors. Automatic segmentation and tumor detection in cervical MR data are addressed in this paper. The proposed method of registration identifies the boundary of an organ of interest based on Hierarchical Adaptive Local Affine Registration. It combines the transformation at each level of all local affine components to form an overall smooth transformation. Optimization is achieved by Hybrid Particle Swarm Optimization (HPSO).The proposed approach may achieve an improved accuracy than the existing registration algorithms such as Rigid Registration and Non-Rigid Registration.