A new arXiv paper presents MedSAM-3, a medical image segmentation model that takes text prompts as input. The authors argue that existing segmentation methods generalize poorly and require extensive, time-consuming manual annotation whenever they are applied to a new clinical task. The work appears as a replacement cross-listing on arXiv (2511.19046v2).
A new arXiv paper proposes CANAL, a method that lets hospitals train segmentation models together without sharing patient scans. It uses knowledge distillation to transfer learned feature representations, injecting differential-privacy noise in a channel-aware way so that the privacy budget is spent where it matters most. The approach targets medical image segmentation, where complementary data sits in separate institutions that cannot legally exchange it.
A new arXiv paper examines where and how much neural networks forget when trained sequentially on gynecological medical imaging tasks. The authors analyze forgetting layer by layer to identify which parts of a segmentation model should be preserved and which can be adapted. The work targets scenarios where consecutive clinical tasks vary in imaging modality and anatomy.
Researchers on arXiv examine whether clinical text inputs genuinely drive pixel-level predictions in pretrained vision-language models for medical image segmentation. They propose an evidence-decoupling approach to characterize the sensitivity of the text branch, aiming to clarify the real contribution of textual information to segmentation outputs.
A new arXiv paper introduces LightMedSeg-ISLES, a 1.26-million-parameter pipeline for segmenting stroke lesions in T1-weighted MRI scans. The authors position it as a lighter alternative to large networks and ensembles like nnU-Net, citing an 81-fold reduction in parameters. The smaller footprint is intended to ease storage and inference demands for clinical deployment.
Researchers propose Referee-Based Quality Estimation, a framework that scores polyp segmentation output without ground-truth labels by measuring how much a primary model agrees with other models. The approach is intended as a deployment-time signal that catches silent failures during real-time colonoscopy, when annotations are not available at inference. The work is published as an arXiv preprint.
A new arXiv paper describes a vision-language reinforcement learning framework for on-demand analysis of medical images across a range of clinical applications. It focuses on delineating tumors and nearby organs at risk, a step central to radiotherapy planning, surgery, and treatment response assessment that currently demands significant expert time. The authors position the method as addressing the limitations of existing AI systems in this domain.