Echo-DM: Ultrasound marker removal via conditional latent diffusion and region-aware fusion

Clinical ultrasound images often contain artificial markers, such as measurement calipers and text, to assist diagnostic interpretation and comparison. However, these markers can introduce shortcut bias in downstream automated analysis, encouraging deep learning models to rely on marker-related cues rather than clinically meaningful anatomy. Existing marker removal methods are either mask-dependent and vulnerable to error propagation, or mask-free deterministic restorers that may over-smooth ultrasound texture and perturb unaffected background regions. To address these challenges, we present Echo-DM, a framework for ultrasound marker removal via conditional latent diffusion and region-aware fusion. Echo-DM follows a generate-select-refine formulation: a DiT-based Conditional Latent Diffusion Network (CLDN) generates an intermediate restoration, and a Region-Aware Fusion (RAF) module internally predicts a soft map to selectively fuse the restoration with intact input content before residual refinement in image space under end-to-end mask-free inference. The image-space RAF is latent-module-agnostic and is shared by Echo-DM-V and Echo-DM-R, which use VAE-based and RAE-based latent modules, respectively. Extensive experiments on Echo-PAIR, a large-scale paired clinical ultrasound dataset, demonstrate superior marker removal and strong anatomical fidelity compared with representative two-stage baselines, while providing favorable quality-efficiency trade-offs across deployment settings. Code and models will be released at Echo-DM.