BrainTumor CNN: brain tumor classification via transfer learning and ResNet-18
TL;DR
This study proposes BrainTumor CNN, a convolutional neural network (CNN) for classifying brain tumor MRI images, which leverages transfer learning via a pre-trained ResNet-18 network, integrating data augmentation and Dropout regularization to enhance robustness and generalization.
Abstract
Brain tumors affect millions of patients worldwide. Magnetic resonance imaging (MRI) is the preferred detection modality due to its radiation-free nature, high soft-tissue contrast, and three-dimensional (3D) imaging capabilities. However, the 3D complexity of MRI scans makes manual classification time-consuming, inefficient, and prone to errors. Consequently, developing high-precision automated classification is crucial in neurology. This study proposes BrainTumor CNN, a convolutional neural network (CNN) for classifying brain tumor MRI images. It leverages transfer learning via a pre-trained ResNet-18 network, integrating data augmentation and Dropout regularization to enhance robustness and generalization. The model demonstrates exceptional diagnostic performance, achieving 98.6% accuracy, with precision, recall, and F1-score all reaching 99.8%. By balancing high accuracy with low computational cost, this study provides an efficient diagnostic tool suitable for real-time clinical deployment.