CT Image Segmentation for Prostate Cancer Diagnosis Based on 3D U-net Deep Learning Model

Kryzsko Commons Ballroom, Winona, Minnesota

The vast advancement in deep learning techniques over the past years has consequently led to the creation of state-of-the-art models applied in the medical imaging field for cancer diagnosis. The basic 3D U-net model has shown success in delimiting various organs on medical images such as CT scans, MRIs, and X-Rays. However, it is desirable to develop a more accurate segmentation model. In this work, we study the basic 3D U-net model applied to CT image segmentation of the male prostate organs for prostate cancer diagnosis and compare its performance with that of an improved 3D U-net model. Experiments are carried out on a dataset of CT images with prostate organs and the performance is evaluated using the Intersection over Union (IoU) metric.