A New Hybrid Region-Based Segmentation for 2D Corpus Callosum Segmentation

The development of an image-based brain image segmentation system using UNet has the advantages of a network that does not contain a fully contained layer. These steps have involved modifying the fully convolutional networks architecture proposed and extending it to work with very few images and more precise Segmentation. UNet produces only a few features. However, Corpus Callosum Segmentation requires high features and detects the edge of the rostrum, the genu, the body, and the splenium to achieve higher performance. This paper proposes UNet ++ with A New Hybrid Region-Based Segmentation (NHRBS) as a new region-based network strategy by combining region-based Segmentation with UNet++ that were improving object detection in 2D Corpus Callosum object segmentation. Our test results show that NHRBS accomplished a dice coefficient of 0.99.