Multi-Class Brain Tumor Classification with VGG Models: An MRI-Based Analysis

Brain tumors are abnormal collections of cells resulting from uncontrolled cell division within the brain that pose a significant threat to human health. Misdiagnosis of brain tumors can result in inadequate medical treatment and lead to a decrease in the survival rate for patients. Traditionally, the critical task of identifying and classifying (multi-class classification) brain tumors has rested with highly skilled neuro-oncologists and neurosurgeons. The increase in the number of cases has added to the difficulty of doctors. Convolutional neural networks (CNN) and computer-aided tumor detection systems are success stories that have significantly advanced the field of machine learning. In contrast to traditional neural network layers, deep neural networks can automatically extract the most significant and reliable features from the input region. In this work, we analyzed two Visual Geometry Group pre-trained models (VGG-16 and VGG-19) to classify brain tumors into four specific categories based on location: pituitary, glioma, meningioma, and the absence of a tumor(no-tumor). Each of these explores transfer learning procedures, including fine-tuning and freezing using magnetic resonance (MR) images of brain tumors. In this study, the fine-tuned VGG-19 model achieves high accuracy and outperforms VGG-16 in terms of accuracy and other metrics. This study can therefore assist medical experts in the accurate classification of brain tumors for effective clinical management.