Fatty Liver Level Recognition Using Particle Swarm optimization (PSO) Image Segmentation and Analysis

Abstract

Fatty liver or liver hepatic glycogen is one of the most common disorders of the liver nowadays. Clinical detection of this disorder by a human expert is increasing as our lifestyle leads us toward this phenomenon. So, making a fast and accurate expert system for fatty liver detection is essential in each clinic and that’s why we intended to make one. Proposed expert system, works based on variety of image processing techniques and algorithms to detect fatty liver and recognize its level by four markers. Four segmentation techniques of Otsu, Watershed, K-Means and Particle Swarm optimization (PSO) are employed to determine disorder level. Performance metrics of Accuracy, F-Score and IoU or Jaccard evaluated the precision of the proposed system. Finally, fatty liver level is calculated based on amount of fat deposits inside segmented image. Experiments are conducted on multiple data sample in high resolution with microscope zoom bigger or equal of 200 which are collected from the internet. All performance metrics and comparisons returned satisfactory results in comparing with traditional methods. Proposed system could achieve average accuracy value of 0.922 for all samples comparing with ground truth data. Additionally, F-Score and IoU performance metrics returned values are 0.872 and 0.907, respectively.