ABSTRACT
Because of the liver tissue, it is difficult and important to diagnose liver diseases. Liver and liver lesions segmentation in medical images with precise automated methods are prerequisites for computer-aided diagnostic systems (CADx). In the liver tissue, benign and malignant tumors are difficult to distinguish because of low contrast. Hence, researchers are trying to provide methods for a more accurate diagnosis. In this chapter, the authors provide an automatic segmentation of the liver and liver lesions in high-contrast CT abdominal images using deep learning and image processing. The deep learning network presented in this chapter is based on the well-known U-Net convolutional neural network powered by the 3-D context information. During the training of the proposed network and U-Net, the authors notice that the network sometimes fluctuates in the early stages of training and abandons the decreasing trend of the cost function and finds an upward jump state.
