Deep Learning Advances in Brain Tumor Classification: Leveraging VGG16 and MobileNetV2 for Accurate MRI Diagnostics

• Published in: 2024 International Conference on Power, Energy, Control and Transmission Systems (ICPECTS) pp. 1 - 6
• Main Authors: S, Alex David, Begum, Almas, Kalilulah, S Ismail, Naveena N, Ruth, K, Rajathi, D, Hemalatha
• Format: Conference Proceeding
• Language: English
• Published: IEEE 08.10.2024
• Subjects: Accuracy; Brain modeling; Brain Tumor Classification; Control systems; Convolutional Neural Networks (CNNs); Data Augmentation; Data models; Data preprocessing; Decision making; Deep learning; F1-Score; Magnetic resonance imaging; Measurement; MobileNetV2; MRI Imaging; Precision; Recall; Reliability engineering; VGG16
• DOI: 10.1109/ICPECTS62210.2024.10780014

These modern deep learning models, such as VGG16 and MobileNetV2 will be utilized in this study to classify Brain tumors from a unique specialized MRI dataset The dataset consists of brain tumor original and augmented images as a two-class (yes/no) classification problem domains, where yes indicates the presence of a tumor while learns no corresponds to its absence. Two models VGG16 and MobileNetV2 support deep convolutional layer operations, well known for their very deep architecture; the other one is optimized over efficiency with depth-wise separable convolutions and is used to evaluate their performance in accurately classifying MRI images. The models use similar data preprocessing and regularization techniques; however, their performance metrics are different. Accuracy, precision, recall, and F1-score are the evaluation metrics for each of the five models to understand how well these differentiate between tumor images versus non-tumor images. This study is designed to improve the ability of brain tumor detection, thereby assisting in more reliable diagnosis and clinical decision-making.