Fuzzy Inference System for Liver Ultrasound Image Classification

This research proposed the hierarchical fuzzy based liver classification. Firstly, the initial clustering of liver tissues is extracted using histogram based FuzzyC-Mean clustering. Sub-images are classified into 3 groups: the Most Significant Group (MSG) (the abnormal liver tissue), the Normal Significant Group (NSG) (the normal tissue), and the Least Significant Group (LSG) (a group outside the probe). However, the initial priority map result is inadequate due to a mixture of the abnormal and normal tissues since the quantity, distribution, and reliability of data are not guarantee. The clustering results serve as initial classification. Then, the expert knowledge is integrated via FIS. In order to validate the initial classification, the fuzzy rules based on the second order statistical texture features of Gray Level Co-Occurrence Matrix (GLCM) are suggested. The classification results show 21.98 % and 24.49 % improvement over the histogram and combined histogram and GLCM based FCM clustering. It has been proven that with the integration of the proposed expert knowledge and FCM clustering, the significant improvement of the liver classification can be achieved.