Analysis of Liver Tumor Segmentation using Deep ResUNet

Liver cancer has been the second most fatal cancer to claim a life in men and the sixth most fatal cancer in women. It would be difficult, expensive, time-consuming, and error-prone to manually go through all the Computed Tomography (CT) slices for thousands or even millions of patients, but an early diagnosis with computed tomography could lead to a high recovery rate. On the 3DIRCADb01 dataset, which contains CT slices for patients as well as masks for the liver, tumours, and other physiological areas,Convolutional Neural Network(CNN) is used to solve all the problems. To automate this process, they desired a reliable method that is both straightforward and precise. A hybrid of the Unet and ResUNet models, the ResUNet model uses residual blocks in place of standard convolutional blocks are used here. By using this the liver is segmented and ROI is extracted using the first cascaded CNN, and the tumours were segmented using ROI from the first CNN. Up to 98% real value accuracy and dice efficiency were reached.