Segment feature classification method for hepatocellular carcinoma detection from CT images

Abstract Hepatocellular Carcinoma (HCC) is a chronic liver cancer triggered by hepatitis B or C identified by abdominal pain and jaundice. Computed Tomography (CT) images are used to analyze the type of HCC as primary or solitary based on the infected ratio. This detection and classification utilize machine learning and correlation datasets for precise diagnosis outcomes. In this article, a Segment Feature Classification Method using Region Normalization (SFCM-RN) is introduced. The proposed method identifies heterogeneity measures between the extracted features under two differentiations: the normal and infected regions. The segmentation process is pursued based on null or fewer heterogeneity measures. The differentiations per region are normalized to identify the infection ratio that increases the classifications. Hence this differentiation to normalization case is iteratively verified using sectional deep learning where the single connected layer performs both normalization to differentiation and vice-versa. The iteration is halted if the maximum infected rate is identified through training inputs. The proposed method improves accuracy, F1-score, recall, and precision by 15.52%, 12.51%, 11.24%, and 11.98% respectively. The proposed method reduces error by 14.51% for the maximum feature set.