Advancements and Innovations in U-Net for Enhanced Medical Image Segmentation: A Review

As one of the most prominent network architectures in recent years within the field of medical image segmentation, U-Net has gained widespread adoption due to its exceptional performance, with various extensions continually emerging. Recently, the rapid progress of U-Net has led to a flood of new methods and uses every year. In this context, it becomes imperative to undertake reviews and encapsulate the latest advancements. Such efforts are vital for offering an updated and comprehensible standpoint to researchers interested in this field. This paper highlights the most recent variants of the U-Net, reviews their principles, and summarizes their successful applications by their optimization approaches in medical segmentation tasks. Furthermore, with the success of Transformers in the field of computer vision, there have been many enthusiastic efforts in recent years to apply self-Attention mechanisms to medical image segmentation. This paper delves extensively into how the Transformer model empowers the evolution of the U-Net framework in the context of medical image segmentation. Moreover, it also investigates specific use cases of U-Net in micro and macro tasks such as nucleus segmentation, lung and its nodule, liver segmentation, and gastric detection, all of which hold profound significance in clinical diagnostics. Based on comprehensive literature investigation and analysis, this study provides a forward-looking perspective on the future advancement and possible challenges that U-Net could confront.