Artificial intelligence assisted operative anatomy recognition in endoscopic pituitary surgery

Area of Science:

• Neurosurgery
• Medical Imaging
• Artificial Intelligence

Background:

• Accurate identification of neurovascular structures surrounding pituitary tumors is crucial intraoperatively.
• Existing methods for anatomical identification can be challenging during surgery.
• AI models for sellar anatomy segmentation have been previously developed.

Purpose of the Study:

• To apply a validated AI model for sellar anatomy segmentation.
• To evaluate the impact of AI assistance on clinician anatomy recognition.
• To assess the potential of AI as an intraoperative decision support tool.

Main Methods:

• Clinicians labelled sellar anatomy on images, with and without AI assistance.
• Performance was measured using mean DICE scores and centroid annotation accuracy.
• Participants included medical students, junior trainees, intermediate trainees, and experts.

Main Results:

• Overall mean DICE score improved from 70.7% without AI to 77.5% with AI (p < 0.001).
• Medical students showed the most significant improvement, with DICE scores increasing from 66.2% to 78.9% (p = 0.02).
• AI assistance enhanced accuracy across all participant groups.

Conclusions:

• AI-powered sellar anatomy segmentation significantly improves clinician recognition accuracy.
• This technology shows promise for augmenting surgical education and training.
• AI has the potential to serve as an intraoperative decision support tool in neurosurgery.