Brain Tumor Segmentation Utilizing Thresholding and K-Means Clustering

The segmentation of brain tumors utilizing magnetic resonance imaging (MRI) is a critical step in medical image processing. This results from the valuable information obtained from MRI images that help the radiologist in brain diagnosis. Consequently, many researchers have suggested different methods to address the problem of tumor segmentation in brain MRI images. This paper proposes a brain tumor segmentation algorithm based on k-means clustering, thresholding, and morphological operations. First, K-means clusters the MRI slice into three segments, then a thresholding step converts the segmented image to black and white to separate the tumor from the non-tumor regions. K-means is utilized here as an intermediate step before thresholding to enhance the performance of the segmentation process. On the other hand, non-brain tissue is removed utilizing morphological operations. Four morphological operations have demonstrated significant improvements in the process suggested in this method, erosion, dilation, closing, and opening. The experiments were implemented on BRATS datasets utilizing high-grade (HGG) and low-grade (LGG) images. The results obtained from the simulated experiments demonstrated the powerful achievements of the suggested algorithm in terms of Dice, Jaccard, and F1 score. Furthermore, the suggested method outperforms a few other techniques when applied to the same images.