Enhancing Cancer Detection in MRI Scans Using Transfer Learning and Data Augmentation with Convolutional Neural Networks

Abstract

This research paper explores the integration of transfer learning and data augmentation techniques with convolutional neural networks (CNNs) to enhance the detection of cancer in magnetic resonance imaging (MRI) scans. Leveraging a pre-trained CNN model, the study applies transfer learning to improve feature extraction and classification accuracy, addressing the challenges posed by limited annotated medical datasets. The paper employs extensive data augmentation strategies, including rotation, scaling, and flipping, to artificially expand the training dataset, thereby reducing overfitting and improving the model's generalization capabilities. The proposed method undergoes rigorous evaluation on a publicly available MRI dataset, demonstrating superior performance metrics compared to traditional CNN approaches. Results reveal a significant increase in detection accuracy, specificity, and sensitivity, with the augmented transfer learning model outperforming baseline models by a notable margin. This study underscores the potential of combining transfer learning with data augmentation to develop robust, generalized CNN models for early and accurate cancer detection in MRI scans, promising a substantial impact on diagnostic techniques and patient outcomes. Future work aims at validating the approach across various cancer types and expanding the dataset diversity to further cement the model’s applicability in real-world clinical settings.