Klasifikasi Tumor Otak Menggunakan Support Vector Machine

Research conducted by Sruti et al. has a weakness that is not implementing the segmentation method with image thresholding. Because the segmentation method is a process that can extract the desired object, besides that the segmentation method is also used to isolate the desired object from an image to perform analysis on the object. In this research, the author will be proposing four different scenarios where each scenario will implementing a different image thresholding segmentation method. The data used in this research were MRI images of brain tumors which contain 253 images and divided into 155 images for a tumorous class and 98 images for a non-tumorous class. From the four scenarios that have been proposed, the result are that Scenario number two that uses Niblack image thresholding method with validation accuracy 0.832, training accuracy 0.9653, precision 0.84, recall 0.83, and f1-score 0.875. It is proven that image thresholding has good performance in extracting objects but also has a weakness where the use of image thresholding always discard information on the image because image thresholding always convert the image to the binary