Revision3d improves 3d medical imaging with recursive self correction
ReVision3D: Attribution-Guided Recursive Self-Improvement for 3D Medical Perception
Computer Vision and Pattern Recognition
Summary
Medical imaging machines sometimes miss important details when looking inside the body in 3D scans. The paper presents ReVision3D, a system that helps these machines get better by identifying where they lose important visual information and suggesting specific improvements. It uses 3D images with detailed notes to guide which parts of the imaging process to fix, improving detection of organs like the liver and kidney even when small problems are present. This recursive approach copies and rechecks the images until better results are found.
What this means in practice
- •For medical imaging engineers: Enhance 3D imaging pipelines by using targeted feedback to improve detection of small lesions in volumetric scans.
- •For radiology software developers: Develop diagnostic tools that refine image analysis iteratively, improving recall rates on organs in abdominal CT scans.
Authors
Ho Hin Lee, Yuyin Zhou, Yannan Yu, Shi Gu, Yifan Wu
Abstract
Recursive self-improvement (RSI) offers a promising path for overcoming the limited visual capability of current medical imaging agents. Yet applying RSI to volumetric imaging remains difficult: failures can arise from acquisition, perception, training recipe, or downstream inference, while self-generated feedback and logged trajectories provide little guidance on which component should change. We introduce ReVision3D, an RSI system that leverages 3D volumes with spatially grounded annotations to determine where visual evidence is lost and recursively improve the corresponding visual capability. A frozen language-model designer proposes revisions to acquisition, perception, training, or inference, while the verifier and system-level objective remain fixed. Our key insight is that an annotated volume forms an exact replay world for view rendering and spatial verification: unvisited views can be rendered on demand, and localized predictions can be checked directly against reference masks. This grounded feedback directs targeted revision, while only changes that improve beyond measured seed noise are retained. Each accepted change triggers renewed attribution, allowing the dominant bottleneck to shift across rounds. On abdominal CT, attribution identifies perception as the dominant remaining limitation. Revising that level enables ReVision3D to achieve 79% liver recall and 83% kidney recall at under 0.4 false positives per patient, outperforming the evaluated frozen multimodal foundation models, with the largest gains on small lesions.