Klasifikasi Citra MRI Tumor Otak Menggunakan Metode CNN dengan Arsitektur ResNet-18 dan Inception-V3

This study examines the performance of two Convolutional Neural Network (CNN) architectures, ResNet-18 and Inception-V3, in classifying brain tumor MRI images with and without data augmentation. The dataset was split into training, validation, and test sets in a 70:20:10 ratio. Data augmentation was employed to enhance model variation and generalization. Experimental results indicate that without augmentation, both models experienced overfitting despite achieving high training accuracy (Inception-V3: 100%, ResNet-18: 99.38%). With augmentation, ResNet-18 exhibited a significant increase in accuracy and f1-score, reaching 97%, while Inception-V3 remained stable at 94%. Evaluation using classification reports confirmed that data augmentation effectively improved generalization capabilities, particularly for the ResNet-18 model. Overall, data augmentation proved to be an effective strategy for enhancing the performance and generalization ability of CNN models in brain tumor MRI image classification, with ResNet-18 demonstrating superior performance compared to Inception-V3 under the same conditions.