“FDR-TransUNet: A New Encoder-Decoder Architecture for Medical Image Segmentation with Improved Performance on COVID-19 Lung and Infection Area Segmentation”

https://pubmed.ncbi.nlm.nih.gov/38113680

This study proposes a new encoder-decoder architecture, FDR-TransUNet, that combines U-shaped and Transformer architectures for medical image segmentation, reducing image feature loss during downsampling and improving spatial information recovery during upsamling, achieving better performance on COVID-19 lung and infection area segmentation tasks.