Healthy and Septal Fibrosis Class Liver Ultrasound Image Detection using Deep-Features, Handcrafted-Features and Grey-Wolf Algorithm

The liver is an important internal organ and any abnormality within the liver is a medical emergency. Timely identification and management of liver abnormalities is crucial; hence, various clinical techniques have been established. Of the liver status assessment process, Liver Ultrasound Imaging (LUI) is among the common ones and the aim of this study is to develop an automatic system to assess liver status and Septal fibrosis (SF) in the LUI dataset. Taking advantage of pre-processed images from Shannon's entropy thresholding, a deep learning analysis using local binary pattern (LBP) is proposed to evaluate liver status and improve the results. The Grey-Wolf Algorithm (GWA) is used in the image preprocessing and feature selection process to improve the results. The experimental studies are carried out on both unprocessed and pre-processed images, and then a deep learning model is selected based on its accuracy. The features are then further reduced by using GWA and subsequently concatenated with the reduced features of LBP. This technique enables to acquire fused information and finally to implement a machine learning (ML) classification that allows to classify healthy and SF by ML classifiers. The results of this investigation confirm that the proposed method can detect 100% accurately lizards in the LUI database.