P-TransUNet: an improved parallel network for medical image segmentation

BMC Bioinformatics

Table 3 Comparisons with the state-of-the-art baselines on the 2018 data science bowl (DSB) dataset

Method	mDice	mIou	Recall	Precision
U-Net [5]	0.7573	0.9077	–	–
PraNet [30]	0.8103	0.7108	0.8062	0.8231
Deeplabv3 [37]	0.8857	0.8367	0.9141	0.9081
U-Net + + [6]	0.8853	0.8906	0.8862	0.8628
ResUNet [38]	0.8991	0.8244	0.9000	0.9084
Attention U-Net [11]	0.9083	0.9103	–	0.9161
TransUNet [15]	0.9178	0.8648	0.9023	0.8936
TransAttUnet [39]	0.9162	0.8498	0.9185	0.9193
DS-TransUNet [20]	0.9219	0.8612	0.9378	0.9124
MSRF-Net [35]	0.9224	0.8534	0.9402	0.9022
FCBFormer [36]	0.9245	0.8727	0.9379	0.9083
Our method	0.9363	0.8875	0.9463	0.9237

Terms: The “–” denotes that the corresponding result is not provided. For each column, the best results are highlighted

ISSN: 1471-2105