Implementation and Comparison of Different Segmentation Techniques for MRI and CT images

EasyChair Preprint 4485

Abstract

 Medical image segmentation offers rich information in clinical applications for accurate diagnosis.  Numerous algorithms in image segmentation techniques have been proposed in the field of computed tomography (CT), magnetic resonance imaging (MRI) for brain tumor detection. The Medical practitioner faces the problem in decision-making and choosing the best segmentation method which will apply to determine the region of interest related to tumor lesion. In this paper segmentation of medical images has been investigated using k-means, level set and region growing techniques in brain tumor images. This study provides an accuracy that compare the strength of each method in order to develop necessary computational algorithms that enhance the analysis of biomedical image in diagnosis and analysis purposes.  Image parameters such as mean square error, peak signal-to-noise ratio, normalized cross correlation, average difference, structural content and the accuracy was assessed to compare across each technique. The accuracy of the k-means, level set and region growing is 92.4%, 86.7% and 88.3% respectively.

Keyphrases: Histogram, K-means clustering, Median Filter, Segmentation, level set, morphological operation, region growing

@booklet{EasyChair:4485,
  author    = {Malaz Mahmooud and Tasabih Ibrahim Alamin and Mohammed Y. Esmail},
  title     = {Implementation and Comparison of Different Segmentation Techniques for MRI and CT images},
  howpublished = {EasyChair Preprint 4485},
  year      = {EasyChair, 2020}}