Contrast-Gated Parameter-Efficient Adaptation of MedSAM for Skin-Tone-Robust Lesion Segmentation

Overview CG-MedSAM investigates contrast-conditioned parameter-efficient adaptation of MedSAM for skin-lesion segmentation across skin tones and imperfect bounding-box prompts. The project introduces residual bottleneck adapters modulated by a contrast signal computed within the prompt region. Research scope The manuscript evaluates segmentation quality, prompt perturbations, skin-tone subgroup performance, and contrast-conditioning ablations on ISIC 2018 and sDDI. Experimental findings and limitations should be interpreted using the manuscript and repository documentation. Supporting materials Project repository This deposit contains the author manuscript sourced from repository commit 1b73ab5b1bffa9a16dcbacacf8bc03a3ec42a433, licensed under CC BY 4.0. Datasets and model weights are not included. No journal acceptance or peer-review status is asserted by this deposit.