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

BMC Bioinformatics

Table 6 Generalizability results of the models trained on CVC-clinicDB and tested on Kvasir-SEG

Method	mDice	mIou	Recall	Pre	Flop
Base(resblock)	0.8713	0.8056	0.8964	0.8665	18.8G
Base + T	0.9061	0.8376	0.9272	0.8993	47.6G
Base + P-transformer	0.9253	0.8652	0.9354	0.9190	35.4G
Base + T + GLF	0.9161	0.8479	0.9304	0.9129	50.8G
Base + T + edge	0.9123	0.8474	0.9296	0.9045	49.8G
Base + P-transformer + GLF + edge	0.9352	0.8893	0.9389	0.9379	42.6G

ISSN: 1471-2105