3D Segmentation of Abdominal Organ Images Using Deep Learning

An essential step in the radiological assessment procedure is the delineation of organ areas in the abdomen, or abdominal organ segmentation. The current status of automated methods for separating abdominal organ areas from computed tomography (CT) and magnetic resonance imaging (MRI) is described in this research. The aim is to automatically identify abdominal organs to help radiologists in their day-to-day diagnosis. Segmenting 3D medical images is a complex task, due to the three-dimensional nature of the data and the diversity of organ structures. Segmentation of abdominal organs is a difficult task because abdominal CT and MRI are severely affected by intensity non-uniformity, organ complexity and low contrast during acquisition. The aim of this work is to use a deep learning-based method to segment the liver on CT images for radiotherapy treatment plans. In this project, we used a dataset containing CT scans of 130 patients (MSD). Liver segmentation was performed using the U-Net model. Dice and Dice Loss were used to quantitatively evaluate this delineation.