Characterization and segmentation of Liver in MR images using Texture Analysis

Abstract

This study was aimed to segment and classify the normal liver in MR image. Using pixel intensity and texture analysis to classify and then segment the liver using the classified image then further classify the segmented liver using textural feature. The MRI system used Philips intera 1.5 Tesla and the data collected randomly from 50 patients out of 200. The classification and segmentation processes were carried out using Interactive Data Language (IDL) program as platform for the generated codes. The results of classification were fed to SPSS software to find the classification score for further classification. The total classification accuracy of the segmented liver with it is associated component was 94.3%, with 100% classification for liver tissue, IVC and, while ligament, portal and hepatic vein was 85.7% for each. The most discernible features were the mean intensity feature and the signal feature. In conclusion the applied algorithm showed a potential success to adopt such procedure in medical image processing