Organ Specific Deep Learning Networks for Anatomical Structure Segmentation

Nirmalya Gayen

Abstract

The accurate identification of specific structures within volumetric data derived from diverse scanning devices holds considerable significance in various biomedical applications. However, existing solutions face several challenges. Primarily, the integration of results obtained from the segmented outputs of multiple organs poses difficulty. Additionally, the detection of tubular topological organs, such as the rectum and sigmoid, which have distinct nomenclatures based on their positions, presents challenges. Furthermore, gender-specific organs may exhibit suboptimal performance in current methodologies. Lastly, the detection of small organs within the context of the entire body remains a challenging task. The primary objective of this study is to achieve accurate segmentation of three specific organs from volumetric CT images. Once the segmentation process is completed, the subsequent aim is to determine an optimal needle insertion path and perform a simulation in the identified regions.

[PDF]