MEDICAL IMAGE ANALYSIS USING DEEP LEARNING: A COMPARATIVE STUDY OF CNN AND VISION TRANSFORMER ARCHITECTURES

Medical image analysis has become an essential component of modern healthcare systems due to the increasing availability of digital diagnostic equipment and large-scale medical datasets. Artificial intelligence, particularly deep learning, has enabled automated interpretation of medical images with accuracy comparable to expert clinicians. Among deep learning models, Convolutional Neural Networks (CNNs) have achieved remarkable success in extracting spatial features, while Vision Transformer (ViT) architectures have recently emerged as powerful alternatives capable of modeling global contextual relationships. This paper presents a comparative analysis of CNN and Vision Transformer models for medical image classification tasks. Experimental evaluations were conducted on benchmark dermatology datasets to assess accuracy, loss, and inference efficiency. Results demonstrate that Vision Transformers achieve superior classification performance, whereas CNNs remain advantageous for real-time medical systems due to lower computational cost. The study highlights future research directions for integrating transformer-based architectures into practical clinical applications.