Adaptive Physics-Based Non-Rigid Registration for Immersive Image-Guided Neuronavigation Systems

Area of Science:

• Neurosurgery
• Medical Imaging
• Image Registration
• Computational Anatomy

Background:

• Image-guided neurosurgery relies on co-registered preoperative imaging (anatomical, functional, diffusion tensor imaging) for safe tumor resection in eloquent brain areas.
• Brain deformation during surgery, especially after tumor resection, challenges the accuracy of preoperative image guidance.
• Non-Rigid Registration (NRR) aims to update preoperative images to match intraoperative anatomy while preserving image quality.

Purpose of the Study:

• To compare the accuracy and performance of several non-rigid registration methods for handling brain deformation during neurosurgery.
• To introduce a novel adaptive method (A-PBNRR) that automatically handles deformation caused by tumor resection.
• To present a new mixed reality system integrating ultrasound, MRI, and CT for improved user experience in image-guided surgery.

Main Methods:

• The study involved 30 glioma surgeries with significant tumor resection volumes.
• An Adaptive Physics-Based Non-Rigid Registration (A-PBNRR) method was developed and applied to register preoperative and intraoperative MRI.
• A-PBNRR was compared against rigid registration, B-Spline non-rigid registration (both in 3D Slicer), and traditional Physics-Based Non-Rigid Registration (PBNRR in ITK) using visual assessment, Hausdorff Distance, and landmark-based metrics.

Main Results:

• A-PBNRR demonstrated over fivefold improvement in registration accuracy compared to rigid and traditional PBNRR, and fourfold improvement over B-Spline methods.
• The A-PBNRR method achieved an average application time of under 2 minutes, indicating clinical feasibility.
• Performance analysis confirmed A-PBNRR's ability to effectively model deformation, including that caused by resection.

Conclusions:

• The A-PBNRR method significantly outperforms existing registration techniques in modeling intraoperative brain deformation, particularly after resection.
• Both the accuracy and speed of A-PBNRR are sufficient for clinical application in the operating room.
• A-PBNRR, combined with a mixed reality system, offers a powerful, cost-effective alternative to current neuronavigation systems.