Brain Tumour Segmentation in MRI Data Using Gray Level Co-Occurrence Matrix

In the medical field, the image segmentation process is very important as the first step to diagnose a disease. Currently, brain tumors are one of the most life-threatening diseases. To detect brain tumors more accurately in less time, many techniques have been proposed using image segmentation. This research will present a brain tumour segmentation method on brain MRI data using GLCM (Grey Level Co-Occurrence Matrix). The input data in this system is the planned MRI data obtained from a hospital and Open Source MRI data in the form of grey level which is calculated using GLCM. Calculation using GLCM which is to find the maximum probability among the grey values of surrounding cells. The result of the GLCM calculation produces texture values (Texture Features). The combination of the intensity value in the MRI data and its texture value as an input parameter in the segmentation process. The method used to segment the tumour is thresholding by identifying contours on the MRI image. Both intensity-based and texture-based images are subjected to the same segmentation method. The segmentation result in this study is the categorization of cell areas affected by tumours and normal cells. To see the performance of the system built in this study using DSC (Dice Similarity Coefficient) and JC (Jaccard Coefficient) presented in the form of a percentage. Tests were conducted with 25 MRI data with two types of tumours namely glioblastoma and meningioma. Evaluation using Dice Similarity and Jaccard methods showed a small difference in accuracy between texture and intensity images. The texture image has an average Dice Similarity of 82.81% and Jaccard of 88.40%. While the intensity image has an average Dice Similarity of 80.35% and Jaccard 87.22%. In conclusion, the GLCMCA method provides slightly higher and more precise segmentation accuracy.