Brain Tumor Classification Based Deep Transfer Learning Approaches

• Published in: 2022 International Conference of Advanced Technology in Electronic and Electrical Engineering (ICATEEE) pp. 1 - 6
• Main Authors: Bouguerra, Oussama, Attallah, Bilal, Brik, Youcef
• Format: Conference Proceeding
• Language: English
• Published: IEEE 26.11.2022
• Subjects: Brain modeling; Brain tumor; Classification; Classification algorithms; Deep Learning; Electric potential; Magnetic resonance imaging; Medical services; Neurosurgery; Transfer learning; VGG16; VGG19
• DOI: 10.1109/ICATEEE57445.2022.10093731

The group Doctors without Borders is always working to improve healthcare in underdeveloped countries. By supporting doctors in the identification of brain tumors, artificial intelligence might significantly help them in their endeavors. A neurosurgeon must examine the MRI in order to detect a brain tumor, and experienced neurosurgeons are fairly scarce in third-world nations. In this study, the benchmark dataset for brain tumor detection is used to classify binary data. Our feature extraction algorithms for classification evaluate deep transfer learning with a small number of training datasets. Popular pre-trained models that perform well in image categorization include the VGG16 and VGG19 deep convolutional neural network architectures. The goal of this effort is to use information about the brain to determine whether a tumor is benign or malignant. In this work, we trained our architectures using three distinct medical image enhancement algorithms with little preprocessing in order to explore the effects on classification performance. Our readings support the idea that transfer learning produces trustworthy outcomes when applied to small datasets. The suggested system achieves 99.77% classification accuracy, outperforming state-of-the-art techniques.