Recent developments in segmentation of COVID-19 CT images using deep-learning: An overview of models, techniques and challenges

• Published in: Biomedical signal processing and control 2024-05, Vol.91, p.105970, Article 105970
• Main Authors: Zhang, Ju, Ying, Changgan, Ye, Zhiyi, Ma, Dong, Wang, Beng, Cheng, Yun
• Format: Article
• Language: English
• Subjects: COVID-19; CT imaging; Deep learning; Medical image segmentation
• ISSN: 1746-8094; 1746-8108
• DOI: 10.1016/j.bspc.2024.105970

The outbreak of the COVID-19 has resulted in a catastrophic situation worldwide and has become one of the most serious diseases in the last hundred years. In recent years, with the rapid development of deep learning technology, deep learning-based segmentation of COVID-19 CT images methods have become quite popular because of their data-driven and high-performance features to achieve excellent segmentation results. However, to our knowledge, no relevant review has so far comprehensively introduced and reviewed advanced deep learning segmentation of COVID-19 CT images such as Transformer structures in segmentation tasks. Therefore, this study presents a systematic survey of current situation of segmentation of COVID-19 CT images, challenges and future research directions based on the literatures related to segmentation of COVID-19 CT images published from year 2016 to 2023, and in particular, the relevant literature from 2020 to 2023. Four categories of segmentation of COVID-19 CT images methods are classified according to the network structure including encoder-decoder based models, attention-based models, multi-scale and pyramid-based models and Transformer based models. Meanwhile, we summarize COVID-19 CT datasets and image segmentation evaluation metrics that are currently publicly available to evaluate and train the networks. Furthermore, data augmentation methods and loss function are also presented. The representative methods are experimentally compared and analyzed and summarized. Challenges and opportunities of deep learning segmentation of COVID-19 CT images in current and future directions are also discussed and presented.