Improving COVID-19 Pulmonary Damage Volume Prediction with Uncertainty Estimation and Prediction Refinement Embedded in Trans-UNet

Accurate prediction of COVID-19 pulmonary damage volume is paramount for diagnosis and prognosis. In this study, we introduce a Trans-UNet-based model that advances the field by incorporating an Uncertainty Prediction branch, allowing for precise prediction confidence quantification. This addition not only elevates prediction accuracy but also facilitates clinical decision-making through attention maps, providing valuable insights. Our research holds promise for the advancement of medical image analysis and the enhancement of COVID-19 patient care.