Automated Brain Tumor Classification from MRI using a pretrained ResNet50 architecture

Abstract

In today’s era, accurately and promptly detecting brain tumors is important for effective patient care. Although MRI is a popular technique for examining a patient’s brain structure and detecting medical abnormalities, manual interpretation can be time-consuming and may vary depending on the individual. This research employed a fine-tuned ResNet-50 pre-trained architecture, which has been pre-trained on the high-volume ImageNet dataset, for the automated classification of four types of tumor images using a recently popular transfer learning technique.

The suggested approach employed 3264 MR images, divided into training and testing data. After analysis of training and testing data, the model achieved 51% accuracy on test data. Furthermore, the outcome highlights the task's complexity and suggests possibilities for improvement in future work. The research work highlights the promise of transfer learning but suggests further optimization, including the application of advanced techniques and a comparative analysis with other pre-trained architectures, to enhance diagnostic accuracy. In addition, future work suggests advanced fine-tuning strategies, regularization techniques, and other methods to enhance model performance, thereby aiding medical professionals in brain tumor diagnosis.