基于KAN与CKAN优化的医学图像分割模型

娄世猛,邵玉斌,杜庆治,唐菁敏,张赜涛

Medical image segmentation model based on KAN and CKAN optimization

Shimeng LOU,Yubin SHAO,Qingzhi DU,Jingmin TANG,Zetao ZHANG

表 3 UNet、KUNet、nnUNet、SWin-UNet模型在各数据集上的最佳性能

Tab.3 Best performance of UNet, KUNet, nnUNet and SWin-UNet model on each dataset

网络模型 Dice IoU LiTS CORN DRIVE Lungs LiTS CORN DRIVE Lungs UNet 0.9809 0.5133 0.8104 0.9609 0.9626 0.3531 0.6813 0.9248 nnUNet 0.9917 0.3219 0.7981 0.9747 0.9837 0.2049 0.6521 0.9563 Swin-UNet 0.9842 0.0815 0.6875 0.9571 0.9688 0.043 0.5239 0.9177 KUNet 0.9905 0.5302 0.9301 0.9842 0.9811 0.3681 0.8694 0.9689 $ \varDelta $ 0.1348↓ 0.0353 0.9225 0.4623 0.1477↓ 0.0235 0.8373 0.3369 $ \overline{\varDelta } $ 0.3213 0.2625