MIT Researchers Build 5-Minute AI Tool for Sub-Millimeter Surgical X-Ray Matching
Updated
Updated · MIT News · Sep 16
MIT Researchers Build 5-Minute AI Tool for Sub-Millimeter Surgical X-Ray Matching
1 articles · Updated · MIT News · Sep 16
Summary
Nature published MIT’s xvr system, which adapts to each patient in about five minutes and then aligns intraoperative X-rays with preoperative CT or MRI scans in seconds.
Sub-millimeter registration aims to make minimally invasive procedures safer and faster by showing clinicians exactly where catheters, endoscopes and other tools sit inside the body.
Xvr generates roughly 1,000 synthetic X-ray images per second from a patient’s 3D scan, using physics-based simulation to train a patient-specific model without relying on broad one-size-fits-all predictions.
More than 2,000 whole-body scans were used to pretrain a foundation model, and tests on data from five hospitals showed xvr beat existing AI methods by an order of magnitude across varied patients and procedures.
The team is now working on real-time deployment, moving-body scenarios and partnerships with surgical robotics and clinical groups, with potential use in emergency interventions such as stroke care.