Domain-continual learning for multi-center anatomical detection via prompt-enhanced and densely-fused MedSAM

Open Webpage

Zhan Gao, Ling Huang, Qika Lin, Bin Pu, Kai He, Mengling Feng, Kenli Li

Published: 01 Jan 2026, Last Modified: 06 Nov 2025Inf. Fusion 2026EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•DCA-Det enables robust fetal structure detection in multi-center ultrasound data.•Leverages MedSAM and two-stage training to improve continual domain adaptation.•Self-prompted decoder guides features to anatomical targets and reduces noise.•Dense fusion module enhances fine-grained detail and cross-scale interaction.