Fat detection algorithm for liver biopsy images

This paper presents an image-processing-based method for analyzing the fat proportion in liver biopsy images. Firstly, the proposed method extracts the area of candidate fat blobs, as well as the background area, from the input image. Then the features of each candidate blobs will be computed. Finally a classification technique called k-nearest neighbors is used to classify each candidate blob if it is fat. Experimental results show that the proposed method can detect fat in the liver biopsy images with the accuracy of 97.52%.