Medical image segmentation using UNet algorithm

Abstract

Medical image segmentation is one of the main processes to diagnosis the disease. Deep learning is the popular domain to segment the medical image. Now days accurate liver segmentation is a main problem because the large shape and unclear boundaries of the abdomen. This research mainly focused on segmenting liver from the abdominal CT scan images using a deep learning method and minimizing the effort and time used for diagnosis. This algorithm is based on the UNet algorithm in deep learning. This work proposed a liver segmentation method using a UNet architecture as a baseline. The number of filters in the network layer is reduce based on the complexity of the segmentation to improve the performance. Loss function in Unet is the main factor for performance of the network. The background imbalance of the images is reduced using loss function. Efficacy of the U-Net was demonstrated using the public dataset 3DIRCADb-01. The Dice Similarity Coefficient (DSC) of the proposed work is 97.76% and the Volumetric Overlap Error (VOE) based on the work is 0.0177. this network will helpful for our practical work.