Fused Second-Order MR Image Feature Framework for Region of Interest Delineation

Abstract Healthcare infrastructure relies on technology-driven solutions such as CAD systems for improving the overall efficiency of its procedures and processes. Image segmentation is one of the most critical phases for such systems in view of the fact that accuracy of this phase determines the efficacy of the later phases, to a large extent. Extensive research is underway to develop segmentation techniques that can achieve highest accuracy with some suggestions directed towards an information fusion based approach within the machine learning paradigm. This research paper proposes a fused second-order statistical image feature framework for Region of Interest delineation. It is a feature fusion-based segmentation approach (ACM-FT) that fuses texture driven feature maps from GLCM , GLRLM and Gabor filters. The proposed approach is then compared with Active Contour Model with classical edge detection method (ACM-ED) and Active Contour Model without edges (ACM-WE) using Overlap Index (OI) and Jackard’s Similarity Co-efficient (JSI). The proposed approach achieves an average accuracy of 92.17% and 93.19% for JSI and OI, respectively, demonstrating significant improvements.