Systematic Review of Swin Transformer Architectures for Medical Image Classification

Abstract

The growing complexity and volume of medical imaging data have driven the need for advanced automated classification systems that combine precision with computational efficiency. Swin Transformers have emerged as a transformative architecture in this domain, leveraging hierarchical feature extraction and shifted window attention to address the unique challenges of medical image analysis. Unlike conventional approaches, Swin Transformers excel at capturing both fine-grained details and broader contextual information, making them particularly suited for tasks requiring multi-scale analysis. This systematic review explores the role of Swin Transformers in medical image classification, examining their architectural advantages, clinical applicability, and current limitations. Future research should focus on optimizing Swin Transformers for real-world healthcare settings, including the development of lightweight variants and robust benchmarking protocols. The review also identifies gaps in current applications, particularly in understudied modalities like ultrasound, and calls for more comprehensive datasets to enhance model generalizability. By addressing these challenges, Swin Transformers have the potential to advance medical image classification significantly, bridging the gap between cuttingedge AI research and practical clinical implementation. This work provides a foundation for further innovation in the field, emphasizing the need for interdisciplinary collaboration to realize the technology's promise in healthcare.

Ähnliche Arbeiten