An endoscopic chisel: intraoperative imaging carves 3D anatomical models.

Area of Science:

• Medical Imaging
• Computer Vision
• Surgical Navigation

Background:

• Preoperative CT scans are crucial for sinus surgery planning and navigation.
• Intraoperative anatomical changes during surgery render preoperative models inaccurate.
• Current navigation systems lack real-time adaptation to surgical modifications.

Purpose of the Study:

• To develop a vision-based approach for updating 3D anatomical models during navigated sinus surgery.
• To leverage intraoperative endoscopic video to correct for anatomical changes.
• To enhance the accuracy and reliability of intraoperative navigation.

Main Methods:

• A vision-based method using intraoperative monocular depth estimation from endoscopic video.
• Comparison of intraoperative depth estimates with preoperative renders to identify modified regions.
• Volumetric fusion using a truncated signed distance function to update the 3D model.

Main Results:

• Quantitative evaluation on an ex vivo specimen demonstrated reduced model error during surgical progression.
• The updated models accurately reflected anatomical changes compared to ground-truth CT.
• The approach showed improved accuracy over models without real-time updates.

Conclusions:

• Preoperative 3D models can be dynamically updated using intraoperative endoscopy video in navigated sinus surgery.
• This method enhances surgical precision by providing a more accurate, real-time anatomical representation.
• Future work aims to improve depth estimation and reduce reliance on external navigation systems, paving the way for a digital twin in sinus surgery.