Brain Tumor Segmentation and Classification using Texture Features and Support Vector Machine
Leah McIntyre, Eva Tuba
Abstract
Brain tumors are masses of abnormal tissue that can grow to become cancerous and can seriously impact an individual’s health. They can be detected on brain MRI images. Given that early detection of brain tumors is highly important for diagnosis and treatment, many computer aided diagnostic systems based on image processing techniques were proposed to aid in tumor detection. Two very important steps in these methods are segmentation and classification. Segmentation methods such as threshold, region-based techniques and pixel classification are used to separated brain tumor tissue from healthy brain tissue on MRI images. Classification techniques, such as support vector machine and artificial neural networks, have been used to classify between normal brain image and image with a tumor, i.e. tumor and no tumor. This paper proposes a method that includes preprocessing brain MRI images and segmentation using k-means clustering algorithm. Texture features are extracted from the region of interest by using the gray level co-occurrence matrix and used to train a support vector machine for classification. With a 95.21% accuracy score, this model was seen to be comparable to other previously proposed models in the literature.