Abstract

Primary adult liver cancer is currently classified into two main diagnostic categories: cholangiocarcinoma and hepatocellular carcinoma. Hepatocellular carcinoma is the most common kind of liver cancer in adults. These are extremely heterogeneous tumours with a wide range of morphological and clinical characteristics, which reflects the wide range of oncological drugs available and the intricate pathways that lead to carcinogenesis. Because of the large quantity of data acquired from phenotypic and molecular research, the classification of liver cancer is shifting away from the old method, which is based on morphological aspects, and toward a more functional approach based on functional characteristics. In this paper, we develop a Capsule Network (CapsNet) classifier to classify the liver regions from computerized tomography (CT) images. The CapsNet helps in classification of instances using pre-processing and feature extraction stages. The training of the classifier is conducted using various liver images from the input datasets and the classifier is validated using the test images. The simulation is conducted to test the effectiveness of CapsNet and the results of simulation shows that the proposed method achieves higher degree of classification than other methods.