Brain Tumor Detection and Classification Using Geometrical Shapes as Texture Descriptors

Main Article Content

Muhammad Zeeshan Zafar
Syed Muhammad Adnan
Junaid Rashid
Wakeel Admad
Javaria Ikram


The existence of abnormal cells in the brain causes a brain tumor. There are two kinds of the tumor; Low grade or slow-growing tumor and High-grade or fast-growing tumor. Patient's successful treatment depends on the accuracy of tumor detection. Therefore, an automatic system with improved accuracy for tumor detection and classification is required. The proposed method consists of three phases to determine the presence of a brain tumor. Red, Green, and Blue (RGB) input image is converted into gray level image and skull is stripped using image masking in the preprocessing phase. In the second phase, the features are extracted using geometrical descriptors. These geometrical descriptors comprise of three geometrical shapes eclipse, parabola, and hyperbola. The performance of these geometrical shape descriptors is evaluated using Local Ternary Pattern (LTPs) and Local Quinary Patterns (LQPs). Support Vector Machine (SVM) and k-Nearest Neighbors (KNN) classifiers are used for classifying MR Images into the healthy and unhealthy brain. Experiments are performed on the Kaggle brain MR Imaging dataset and results are compared with existing techniques. Our experimental results show that the parabola descriptor achieved 97.5% as compared to other geometrical shape descriptors like eclipse and hyperbola.

Article Details

How to Cite
Zafar, M. Z., Adnan, S. M., Rashid, J., Admad, W., & Ikram, J. (2019). Brain Tumor Detection and Classification Using Geometrical Shapes as Texture Descriptors. Technical Journal, 24(01), 83-89. Retrieved from