Vision Transformers for Biomedical Applications *

Advanced artificial intelligence (AI) methods, such as attention mechanisms using Transformer architecture, typically used for large language models, have proven their potential in biomedical applications. ViT is proven to be successful in various biomedical applications, such as cancer detection, remote photoplethysmography (rPPG), COVID-19 detection, and lung disease detection. As such, this paper presents a systematic review of the research gaps and future directions of Vision Transformers (ViT)s in biomedical applications to investigate their efficiency in integrating transformers with computer vision applications in the biomedical domain. The paper also explores the lightweight Mobile Vision Transformer (MoViT) models and their applications in the biomedical field. The reported outcomes of applying ViTs in detecting diseases, estimated heart rates and surgical applications are promising and have the potential to be transformed into treatment monitoring and medical diagnosis. It is also worth highlighting that ViT has demonstrated promising results in COVID-19 detection. In future work, it is recommended to improve the performance on ViT through adapting pre-trained models, the addition of more convolution layers, and hyperparameter optimisation. MoViT is a promising lightweight, advanced AI model that is useful in real-time computer vision applications in the biomedical field. ViTs are promising for improving medical research.