A BRIEF REVIEW OF SEGMENTATION METHODS FOR MEDICAL IMAGES

V. S. Khandekar, K. A. Samani

Abstract: For medical diagnosis and laboratory study applications we cannot directly use image that are acquired and detect the disorder because it is not efficient and unrealistic. These images need processing and extracting portions from them that can be used for further study or diagnosis. The main goal of this paper is to give overview about segmentation methods that are used for medical images for detecting the edges and based on this detection the disease prediction and diagnosis is done. There are a lot of tools available for this purpose such as STAPLE and FreeSurfer whole brain segmentation tool etc. Some of these methods are semiautomatic i.e. they require human intervention for their completion and some of them are automatic. The methods are totally divided into four types namely, edge based segmentation, region based segmentation, data clustering and matching. The aim of segmenting medical images is that to detect the ROI and diagnose for a disease based on the detected part. Segmentation is partitioning a image into meaningful regions based upon a specific application. Generally segmentation can be based on the measurements like gray level, color, texture, motion, depth or intensity. Segmentation is necessary in two situations, namely, setoff segmentation i.e. when the object to be segmented is interesting in itself and can be used separately for further studies, and secondly concealing segmentation i.e. suppose there are some noise or vision blockers in the image, so this segmentation aims at deleting the disturbing elements in an image. This paper focuses only on the working of different methods that are used for segmentation whether they segment well or poor.

Keywords: Image Registration, Image Segmentation, Reinforcement Learning,

DOI: https://doi.org/10.15623/ijret.2014.0301073