Liver Cirrhosis Classification on M-Mode Ultrasound Images by Higher-Order Local Auto- Correlation Features

Ultrasound images are widely used for diagnosis of liver cirrhosis. In liver cirrhosis classification using Mmode ultrasound images, Zhou’s method has been shown to be effective. However, in Zhou’s approach, the liver cirrhosis classification performance depends on the accuracy of the abdominal aorta wall extraction. Therefore, we examine to classify the liver cirrhosis not using the abdominal aorta wall extraction. In this paper, we propose a liver cirrhosis classification method using higher-order local auto-correlation (HLAC) features. Furthermore, we propose to use image processing techniques of a thresholding technique and a shading technique to effectively extract the HLAC features. We also examine a feature selection method by Fisher ratio to reduce the dimensionality of the HLAC features. Experimental results show the proposed method is promising. The average error rate of the proposed method achieves 12.11(%).