ANALISIS KINERJA ARSITEKTUR CNN ALEXNET DAN VGG16 UNTUK KLASIFIKASI TUMOR OTAK

Abstract

Early detection of brain tumors is essential for determining appropriate treatment strategies and increasing patient survival rates. This study analyzes and compares the performance of two Convolutional Neural Network (CNN) architectures Alexnet and VGG16 for classifying brain tumor MRI images into three categories: glioma, meningioma, and pituitary. The dataset, annotated by medical experts, was split into 80% for training and 20% for testing. Each image underwent preprocessing steps including resizing, normalization, and data augmentation. Both models were initialized with pre-trained weights from ImageNet and trained for 15 epochs using the Adam optimizer. Model performance was evaluated using accuracy, precision, recall, and F1-score metrics. The results show that Alexnet achieved a testing accuracy of 78.99% with a weighted F1-score of 0.79, while VGG16 obtained an accuracy of 78.01% and a weighted F1-score of 0.75. Although VGG16 has a deeper architecture capable of capturing more complex features, Alexnet demonstrated more stable and balanced performance across all tumor classes. These findings suggest that Alexnet is more effective for classifying brain tumor MRI images within the evaluated dataset and holds strong potential for integration into medical decision-support systems based on deep learning.