Skip to main content
[Preprint]. 2023 Feb 2:2023.01.31.23285223. [Version 1] doi: 10.1101/2023.01.31.23285223

Table 3.

Performance of models on DDTI and Stanford CINE datasets. Bolded values highlight the model with the best performance on a certain metric.

DDTI
Stanford CINE
Average Across Positive Images
Average Across Positive Images
Model Name DSC + IoU DSC + IoU
UNet-AD 0.494±0.242 0.364±0.223 0.456±0.264 0.335±0.236
NestedUNet-AD 0.502±0.247 0.372±0.227 0.476±0.275 0.356±0.244
SResUNet-AD 0.576±0.224 0.438±0.212 0.535±0.251 0.404±0.230
SGUNet-AD 0.513±0.257 0.384±0.232 0.530±0.271 0.405±0.248
AttUNet-AD 0.504±0.234 0.370±0.215 0.462±0.264 0.339±0.232
MSU-Net-AD 0.527±0.325 0.421±0.290 0.569±0.322 0.463±0.294

UNet 0.639±0.203 0.501±0.211 0.616±0.245 0.488±0.242
NestedUNet 0.614±0.195 0.470±0.195 0.609±0.258 0.483±0.249
SResUNet 0.604±0.214 0.466±0.215 0.601±0.251 0.472±0.242
Wang et al. 0.728±0.241 0.617±0.243 0.701±0.303 0.609±0.299
SGUNet 0.587±0.202 0.443±0.197 0.563±0.253 0.433±0.234
AttUNet 0.649±0.191 0.509±0.204 0.612±0.244 0.483±0.240
MSU-Net 0.705±0.191 0.575±0.210 0.658±0.269 0.542±0.267