Real-Time 3-D Video Reconstruction for Guidance of Transventricular Neurosurgery.

Area of Science:

• Neurosurgery
• Medical Imaging
• Robotics

Background:

• Deep-brain neurosurgery using endoscopic approaches can cause ventricular and parenchymal deformation, compromising conventional neuronavigation accuracy.
• Accurate real-time guidance is crucial for targeting deep-brain structures, especially when anatomical distortions occur during procedures.

Purpose of the Study:

• To develop and evaluate a 3D endoscopic reconstruction and registration method using simultaneous localization and mapping (SLAM) for real-time neurosurgical guidance.
• To enable augmented video overlay of preoperative or intraoperative 3D images within the endoscopic view for enhanced targeting accuracy.
• To assess the method's performance and geometric accuracy under challenging conditions, such as limited data and visual scene occlusion.

Main Methods:

• A novel method for 3D endoscopic reconstruction and registration employing SLAM was developed.
• Phantom studies were conducted to evaluate geometric accuracy and uncertainty in scenarios with feature sparsity and scene occlusion.
• Performance was assessed under various data limitations, including up to 40% feature density loss and 120° visual scene occlusion.

Main Results:

• The SLAM-based method demonstrated high geometric accuracy, achieving a target registration error of 1.02 mm.
• Reconstruction and registration accuracy were maintained even with significant feature loss or visual occlusion.
• The system achieved real-time guidance capability with a runtime of 3.45 Hz, representing a >16x speedup compared to previous methods.

Conclusions:

• The developed SLAM-based 3D endoscopic reconstruction and registration method provides accurate, real-time guidance for deep-brain neurosurgery.
• The technique effectively compensates for anatomical deformation and challenging data conditions, enhancing targeting precision.
• Quantitative validation establishes the method's potential for future clinical translation in neurosurgical navigation.