Deep Learning Techniques in Liver Segmentation: Evaluating U-Net, Attention U-Net, ResNet50, and ResUNet Models

When compared to other cancers, liver cancer has a noticeably higher death rate, making it stand out. The necessity for an automated liver segmentation model to support doctors in the accurate identification and treatment of liver lesions becomes clear when considering how laborious and prone to error manual liver lesion segmentation is. Determining and accurately delineating the borders of the liver using medical images such as CT scans or MRIs is a critical challenge in liver segmentation. Accurate liver segmentation is essential for determining its volume, identifying liver abnormalities, and facilitating computer-assisted diagnosis of liver illnesses. It is used for monitoring illness progression, diagnosing diseases, and planning surgical procedures. It makes it easier to precisely and painstakingly define the liver region in medical imaging and provides important information on the shape, size, and location of the liver. To alleviate the time constraints and address the inherent variability associated with manual segmentation, automated liver segmentation methods have been devised to support medical professionals in this endeavor. To tackle this issue, we performed liver segmentation using different models like ResNet, UNet, Attention UNet, and ResUNet.