Fig. 7: GenSeg’s end-to-end data generation mechanism significantly outperformed baselines’ separate generation mechanism.
From: Generative AI enables medical image segmentation in ultra low-data regimes
a The in-domain generalization performance of GenSeg, which performs data generation and segmentation model training end-to-end, compared to the Separate baseline, which performs the two processes separately, when used with UNet or DeepLab in segmenting placental vessels, skin lesions, polyps, intraretinal cystoid fluids, foot ulcers, and breast cancer utilizing the FetReg, ISIC, DermQuest, CVC-Clinic, KVASIR, ICFluid, FUSeg, and BUID datasets. b GenSeg’s out-of-domain generalization performance compared to the Separate baseline in segmenting skin lesions (using examples from ISIC for training, and DermIS and PH2 for testing) and lungs (using examples from JSRT for training, and NLM-SZ and NLM-MC for testing), with UNet and DeepLab as the backbone segmentation models. In all panels, bar heights represent the mean, and error bars indicate the standard deviation across three independent runs with different random seeds. Results from individual runs are shown as dot points. Source data are provided as a Source Data file.
