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

BMC Bioinformatics

Table 1 Comparisons with the state-of-the-art baselines on the Kvasir-SEG dataset

Method	mDice	mIou	Recall	Precision
DoubleU-Net [19]	0.8130	0.7330	0.8400	0.8610
ResUNet++ [31]	0.8133	0.7927	0.8774	0.7064
U-Net [5]	0.8180	0.7460	0.6306	0.9222
FCN [32]	0.8310	0.7370	0.8350	0.8820
DDANet [33]	0.8576	0.7800	0.8880	0.8643
FANet [34]	0.8803	0.8100	0.9060	0.9010
U-Net++ [6]	0.9032	0.8473	0.8923	0.8945
TransUNet [15]	0.9130	0.8570	–	–
DS-TransUNet [20]	0.9130	0.8592	0.9360	0.9164
MSRF-Net [35]	0.9217	0.8914	0.9198	0.9666
FCBFormer [36]	0.9235	0.8757	0.9301	0.9306
Our method	0.9352	0.8893	0.9389	0.9379

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

ISSN: 1471-2105