An Improved Transfer Learning-Based Model with Data Augmentation for Brain Tumor Detection

Current advances in deep learning have brought various breakthroughs in processing medical data. However, dealing with a limited number of medical datasets remains a challenge in deep learning and often leads to overfitting. To solve this research gap, here we show a new approach to improve the performance of a transfer learning-based model for brain tumor detection from 253 brain magnetic resonance imaging (MRI) sample images. The concept of transfer learning has been applied using a pretrained Inception V3combined with data augmentation. Modified layers using dropout and regularization have been additionally utilized to deal with overfitting. The proposed method shows an increase in accuracy of 6.430%, a precision of 5.531 %, a recall of 10.545%, and an F1-score of 8.040% compared to the baseline method. We show that our proposed method has been able to effectively enhance performance and reduce overfitting, even with a small number of datasets. Moreover, our proposed method outperforms state-of-The-Art brain tumor detection.