Impact Of Data Augmentation And Layer Freezing On CNN- Based Brain Tumor Classification Using MRI Scans
DOI:
https://doi.org/10.64252/cqe1ym29Keywords:
Brain Tumor, Convolution Neural Network, VGG16 Model, VGG16 Frozen Layer, Data Augmentation, Performance Evaluation.Abstract
Brain tumors pose significant diagnostic challenges and demand efficient, automated solutions. This research investigates the application of Convolutional Neural Networks (CNNs), focusing on the VGG16 architecture and its frozen-layer variant, for brain tumor detection from MRI scans. A total of 7,023 MRI images were utilized, preprocessed using the Multivariate Fast Iterative Filtering (MFIF) technique, and further enhanced through data augmentation methods including rotation and flipping. The models were trained with the Adamax and SGD optimizers, set to learning rates of
0.001 and 0.0001, respectively. The frozen-layer VGG16 model attained a training accuracy of 99.98% and a test accuracy of 98.12%, along with AUC scores of 1.0 and 0.9994. Evaluation metrics such as Accuracy, Precision, Recall, F1-score, and ROC analysis confirmed that both data augmentation and layer freezing contribute significantly to improving classification outcomes, with an overall average accuracy of 98%.
