Boosting cross-domain semi-supervised medical image segmentation with internal and external regularizations

• Existing SAM-based medical image segmentation methods need high cost, manual prompt, and low inference time. • Activation-restricted internal regularization restricts the influence of noisy pseudo labels by restricting the wrong response value. • SAM as external regularization to advance the generalization of the method. Cross-domain medical image segmentation has been challenging due to the extraordinary cost of collecting sufficient data in various imaging conditions. Mainstream methodologies enhance generalizability by directly adapting large-scale vision foundation models for medical image segmentation tasks. However, these approaches frequently incur high memory costs and necessitate additional prompts for deployment. In this study, we utilize dark knowledge in pretrained segmentation models as external regularization to improve the model’s generalizability. Furthermore, an activation-restricted regularization item is proposed to eliminate the noise/errors within the generated pseudo labels. Experiments on medical cross-domain datasets demonstrate the SOTA performance (above 2.66% improvement) of the proposed method with no additional cost during inference.