A non-invasive technique for diagnosis and identification of liver cirrhosis based on ultrasound images and texture features will be presented in this paper. The proposed system first accepts the acquired US liver image, extracts features, and thereafter, classifies the liver tissue of an input image. The proposed method is based on selecting a variable number of dynamic-sized Region of Interests that are distributed within the liver tissue. The multiple Region of Interests (ROIs) are selected to reduce the required computations. From each ROI, efficient six features (homogeneity, variance, correlation, entropy, contrast, and energy) are selected and extracted. Finally, applying a voting-based classifier using multiple Region of Interests. The proposed method optimizes all the parameter thresholds automatically using the genetic algorithm. The recognition accuracy of this investigation is of 90%.
IntroductionLiver diseases are among the most life-threatening diseases global. Among the different categories of liver lesions, liver cancer has an excessive incidence rate and abnormal mortality rate in Melanesia, East Asia, the countries of Southeast Asia, and sub-Saharan Africa [1]. Liver cirrhosis (LC) represents the end phase of many forms of chronic hepatitis (CH) of different etiologies and with quite variable courses [2]. In chronic liver disease (CLD), the wound-healing process progresses toward