How Efficient is the k-means Clustering to Analyze the CT images of Pyogenic and Amoebic Liver Abscess?

Abstract

Liver abscesses are well-delineated pus-filled lesions. Two common types are Amoebic liver abscesses (ALA), caused by protozoa called entamoeba histolytica while several pus-forming bacteria cause pyogenic liver abscesses (PLA). Both cause debilitating morbidities and are diagnosed by pus culture-sensitivity tests. A contrast CT abdomen shows well-demarcated lesions in the liver. The telemedicine practice is on the rise where image processing is becoming a part and parcel of teleradiology to fill the gap between the number of radiologists versus the large patient pool. Cluster-based image segmentation is a useful step in grouping the image into the desired number of clusters. The k-means clustering (k-MC) technique is one popular method, used in this study on ALA and PLA contrast CT images. it observes that with the desired 2-clusters – a) normal liver tissue and b) the pus-filled tissue) parameters, the algorithm gives better results in PLA.