Abstract
Automated medical image classification is critical for early diagnosis and clinical decision-making. This study evaluates the performance of three deep learning architectures CNN, ResNet, and EfficientNet on benchmark medical imaging datasets including X-ray, CT, and MRI images. Experimental results demonstrate that while CNN achieves baseline accuracy, ResNet improves classification performance through residual connections, and EfficientNet provides the highest accuracy with optimized computational efficiency. Metrics such as accuracy, F1-score, training time, and number of parameters were used for evaluation. Findings indicate that EfficientNet is the most suitable model for practical medical applications requiring both high accuracy and computational efficiency.