MIT News breed · Science

AI Technique Enhances Surgical Precision

Researchers developed an AI technique, xvr, that matches intraoperative X-rays with preoperative 3D scans, potentially improving minimally invasive surgery safety and precision.

Minimally invasive surgeries use X-rays to guide tools, but their flat nature makes precise localization difficult. Manual alignment of X-rays with 3D scans is time-consuming.

The new xvr system adapts to individual patients in about five minutes, automatically matching X-rays with 3D scans in seconds with sub-millimeter precision.

Xvr significantly outperformed existing AI methods across various patients and procedures, potentially making complex procedures more accessible.

The system generates patient-specific, physics-based synthetic X-rays from a 3D scan to train an AI model for accurate 2D/3D image alignment.

A foundation model, pretrained on over 2,000 patient scans, adapts to new patients in about five minutes while maintaining high accuracy.

Testing showed xvr surpassed other AI methods in accuracy and robustness, operating fast enough for emergency surgeries and potentially enhancing robotic surgery.

Future work aims to accelerate xvr for real-time use, validate its reliability, and extend its capabilities to complex scenarios like moving body parts.

AI-samenvatting op basis van de bron.