Characterization of Kidney Infection in Ultrasound B-mode Images Using Texture Analysis

Abstract

The general objective of this study was to develop an algorithm that can extracted textural features from ultrasound images of normal and abnormal kidneys in order to classify these images as having normal tissues, glomerulonephritis, or pyelonephritis. Linear discriminant analysis was used to classify the extracted features from the medulla and pelvic calycle system of kidneys ultrasound images. The results of the study showed that the overall accuracy using medulla texture equal to 98% while for those extracted from pelvic calycle system was 95.7. In conclusion linear function was developed to classify other ultrasound images with an error <5%.