Hepatocellular Carcinoma Diagnosis Based on Ultrasound Images Using Feature Selection Techniques and K-nearest Neighbor Classifier

Abstract

Background: Liver cancer is one of the most common types of cancer, in which early detection plays a significant role in preventing progression and reducing mortality. Ultrasound is one of the methods of liver examination recommended by guidelines due to its performance in detecting focal liver lesions. These small lesions may be missed in the early stages or diagnosed only when the prognosis is poor. Objectives: This study aimed to implement the best classification model for two liver stages by extracting optimal feature subsets to be used in computer-aided diagnosis systems (CAD). Methods: The model classifies the liver into two stages using B-mode ultrasound images of the liver. It involves extracting statistical texture features utilizing discrete wavelet transform (DWT) and gray level co-occurrence matrix (GLCM). This study applied two feature selection methods: t-test and sequential forward floating selection (SFFS). The subset of selected features was presented to the k-nearest neighbor classifier for incorporation into a CAD system. Results: The accuracy, sensitivity, and specificity of the k-NN classifier were 98.75%, 98.82%, and 99.1%, respectively. Conclusions: Image analysis approaches were successfully performed to extract and select useful features. Therefore, this model is recommended for classifying two liver stages, normal and HCC.