An Analytical Comparison of Transfer Learning Techniques for Brain Tumor Detection and Classification

Abstract

Because brain tumors are complex, they must be diagnosed accurately and early. Brain tumor diagnosis requires expert radiologists and large datasets of annotated images, which is resource-intensive and time-consuming. This study investigates the use of transfer learning techniques to detect and classify brain tumors using MRI images. Using transfer learning, pre-trained models can be tuned to perform specific medical tasks while reducing the impact of large datasets. AlexNet, GoogleNet, ResNet-50, and VGG-16 were compared for classifying gliomas, meningioma’s, and pituitary tumors based on transfer learning models. Based on the findings, the proposed hybrid GN-AlexNet model showed superior accuracy, sensitivity, specificity, and F1 score when compared with all other models, demonstrating its potential for improving brain tumor detection efficiency. The findings of this research pave the way for the adoption of transfer learning in clinical settings, providing medical professionals with more efficient and accessible solutions.