Accuracy of thoracic nerves recognition for surgical support system using artificial intelligence

Area of Science:

• Medical technology
• Surgical innovation
• Artificial intelligence in medicine

Background:

• Lung cancer surgery requires precise identification of microanatomies.
• Current visualization methods may have limitations in real-time surgical guidance.
• AI offers potential for enhanced intraoperative visualization and decision support.

Purpose of the Study:

• To evaluate the accuracy of an AI-powered surgical support system in recognizing thoracic nerves during lung cancer surgery.
• To compare the performance of the AI system against traditional surgical monitors.
• To assess the system's usability and integration into the surgical workflow.

Main Methods:

• Development of deep learning models for nerve recognition using annotated surgical images.
• Computational evaluation using Dice and Jaccard indices for accuracy assessment.
• Clinical evaluation by four general thoracic surgeons across 10 thoracoscopic lung cancer surgeries, assessing nerve recognition accuracy, time lag, image quality, and motion smoothness.

Main Results:

• Computational evaluation showed favorable Dice (0.56) and Jaccard (0.39) indices.
• Surgeon evaluation indicated satisfactory accuracy for thoracic nerve recognition (recall: 4.5±0.4, precision: 4.0±0.9).
• The AI system demonstrated minimal time lag (4.9±0.3) and high image quality (4.6±0.5), comparable to surgical monitors, with slight differences in motion smoothness (3.2±0.4).

Conclusions:

• The developed AI surgical support system demonstrates satisfactory accuracy in recognizing thoracic nerves.
• The system shows potential as a valuable tool for enhancing surgical precision in lung cancer operations.
• AI-driven visualization can augment surgeon's ability to identify critical structures during thoracic procedures.