Liver Segmentation Using Hybrid UNet and ResNet-Based Deep Learning Model

Medical imaging plays a vital role in disease diagnosis and treatment. Recently, automated segmentation of images using deep learning models gathers researchers’ attention. The liver is the major largest organ present in the area of the abdomen for the primary or secondary development of a tumor. Recent studies reveal that the automated segmentation of liver lesions is always a challenging task given the CT scan images of the abdomen. Considering this issue an automated hybrid deep learning-based model combining ResNet and UNet had been proposed to perform liver segmentation given CT scan images of the liver covering the region of the abdomen. The performance metrics such as dice coefficient and accuracy were used to predict the performance of the proposed model compared to other existing approaches. Moreover, the segmentation of images is an important technique to perform tasks related to computer vision such as boundary identification and medical image analysis. The experimental results reveal that the proposed model produces accurate segmentation results when compared to other state-of-art methods. The proposed model helps physicians for preplanning their critical surgery and the model shall be fine-tuned and used for the segmentation of other organs.