<![CDATA[Deep learning approaches, particularly Convolutional Neural Networks (CNNs), have proven effective in medical image processing. Two prominent CNN architectures are VGG16 and ResNet18. Previous research has shown that ResNet50 and VGG16 have been used in brain tumor classification with varying accuracy. This study aims to systematically compare the performance of ResNet18 and VGG16 in brain cancer MRI image classification, considering accuracy, computational efficiency, and model generalization capabilities. The results show that the ResNet18 model with pretrained weights achieved the highest accuracy of 97.43%, excelling in detecting all categories of brain tumors with minimal error. In contrast, the VGG16 model trained from scratch performed the lowest with an accuracy of only 63.09%, having significant difficulty distinguishing between classes.]]>
Copyrights © 2025
