An Effective MRI Brain Image Segmentation using Joint Clustering (K-Means and Fuzzy C-Means)

Abstract

This study proposes a segmentation method in human brain MRI images by using a combination of two K-Means and Fuzzy C-Means (FCM) grouping methods to detect brain tumors. K-Means can detect optima and local outliers well and quickly because it is more sensitive to color differences. But the results of the K-Means cluster can be different each time the program starts. To overcome this problem, the results of K-means are clustered again with FCM to classify the convex shape based on the edge so that the cluster results better and the calculation process becomes lighter. Morphology and noise removal processes are also proposed at the preprocessing stage to improve accuracy. In this way the detection results are more effective and accurate with a faster calculation process. Based on the experimental results on 62 brain MRI images obtained an accuracy of 91.94%. This result is far more accurate than the K-Means or FCM methods and also the reverse FCM-K-Means method.