VoxelSage integrates 3D CT analysis with liver tumor surgery planning
VoxelSage: Tool-Augmented 3D CT Analysis and Simulator-Shielded Sequential Resection Planning for Liver Tumors
Computer Vision and Pattern Recognition
Summary
Assessing liver tumors before surgery is complicated because it involves many steps like outlining the tumor in scans, measuring tumor size, and planning how to remove it safely. The authors present VoxelSage, a software system that combines these steps into a single tool using advanced computer methods. It separates language tasks from precise measurement tasks to avoid mistakes and uses a simulator to check surgery plans for safety. The system helps reduce simulated surgery time and blood loss when planning tumor removal.
What this means in practice
- •For radiologists and surgeons: Use an integrated tool to visualize, analyze, and plan liver tumor removal from CT scans with simulation feedback to improve surgical strategy.
- •For medical software developers: Incorporate a dual-port architecture separating language understanding from image processing for robust CT-based tumor analysis and planning systems.
Authors
Binghong Qian, Xuanhe Liu, Yifan Xing, Wenjie Deng, Jian Wu, Haochao Ying
Abstract
Preoperative liver-tumor assessment requires segmentation, physical-space measurement, visual evidence, and resection planning from the same three-dimensional CT volume. Existing tools often handle these steps separately, while language models cannot reliably compute physical measurements from CT. To provide an integrated workflow, we present VoxelSage, a multi-modal system for two- and three-dimensional visualization, liver-tumor analysis, and preoperative resection planning. Its dual-port architecture separates language-model orchestration from image computation: Port A interprets requests and selects skills, while Port B applies them to CT volumes and segmentation masks and returns structured results. Keeping physical measurements in Port B prevents the LLM from computing them directly and reduces the risk of fabricated numerical results. Eight built-in skills support quantitative analysis, visual evidence generation, three-dimensional reconstruction, segmentation refinement, and sequential resection planning; user-defined skills can extend these functions. For sequence planning, a behavior-cloned neural ranker orders candidate resection targets, while a simulator-based shield checks them against predefined constraints. Across 256 unseen simulator scenes, this approach reduced mean simulated time from 34.274 to 33.388 min (0.886 min, 2.59%) and mean simulated blood loss from 300.847 to 183.852 mL (116.995 mL, 38.89%) relative to a deterministic baseline. These results demonstrate system integration and simulator-level performance, not clinical efficacy or safety. The public implementation is available at https://github.com/ZJUMAI/VoxelSage.