# From CNN to Transformer: A Review of Medical Image Segmentation Models

**Authors:** Wenjian Yao, Jiajun Bai, Wei Liao, Yuheng Chen, Mengjuan Liu, Yao Xie  
**arXiv:** [2308.05305](https://arxiv.org/abs/2308.05305)  
**Date:** 2023-08-10  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2308.05305v1/

## Key points

- Medical image segmentation is crucial for accurate disease diagnosis and treatment planning.
- Deep learning techniques, particularly U-Net and transformer-based models like TransUNet, are widely used for medical image segmentation.
- The paper "From CNN to Transformer: A Review of Medical Image Segmentation Models" by Yao et al. surveys four prominent medical image segmentation models.
- The study evaluates the theoretical underpinnings and performance of these models on benchmark datasets such as Tuberculosis Chest X-rays and ovarian tumors.
- Researchers aim to provide insights for developing tailored medical segmentation models specific to different anatomical regions.
- The review highlights advancements in deep learning techniques for medical image segmentation and addresses key challenges and future trends in the field.

## Summary

Medical image segmentation is a crucial aspect of medical image analysis that plays a vital role in accurate disease diagnosis and treatment planning. In recent years, deep learning techniques have gained significant traction in this field, with U-Net and its variants being widely adopted for image segmentation. Additionally, transformer-based models like TransUNet have shown promising results by utilizing pre-trained models from natural language processing tasks. In their paper titled "From CNN to Transformer: A Review of Medical Image Segmentation Models," authors Wenjian Yao, Jiajun Bai, Wei Liao, Yuheng Chen, Mengjuan Liu, and Yao Xie conduct a comprehensive survey of four prominent medical image segmentation models. The study delves into the theoretical underpinnings and performance evaluation of these models on benchmark datasets such as Tuberculosis Chest X-rays and ovarian tumors. By analyzing the characteristics and quantitative outcomes of these models, the researchers aim to provide valuable insights for researchers in the field seeking to develop tailored medical segmentation models specific to different anatomical regions. This review not only highlights the advancements in deep learning techniques for medical image segmentation but also sheds light on key challenges and future trends that could shape the landscape of this critical area of research. Overall, this study contributes to advancing our understanding of state-of-the-art medical image segmentation methodologies and paves the way for further innovations in this vital aspect of healthcare technology.