AI-based visualization of loose connective tissue as a dissectable layer in gastrointestinal surgery

Area of Science:

• Medical technology
• Surgical innovation
• Artificial intelligence in medicine

Background:

• Gastrointestinal surgery requires precise identification of anatomical layers.
• Distinguishing loose connective tissue can be challenging, potentially leading to errors.
• Real-time intraoperative guidance is crucial for surgical safety.

Purpose of the Study:

• To develop an AI model for real-time recognition and display of loose connective tissue.
• To evaluate the performance and clinical feasibility of the AI model in gastrointestinal surgery.
• To assess the impact of AI-driven visualization on surgical judgment and stress.

Main Methods:

• AI model development using training data curated by gastrointestinal surgeons.
• Visual evaluation of model performance on test images and videos by 10 external surgeons.
• Quantitative assessment using Dice coefficient and detection accuracy.
• Qualitative assessment of visualization improvement and surgeon stress levels.

Main Results:

• The AI model achieved a mean Dice coefficient of 0.46.
• Accurate detection of loose connective tissue in 91.8% of images (75% detection threshold).
• False positives were observed but generally deemed non-significant for surgical judgment.
• Surgeons reported improved visualization in 99% of AI-assisted videos with acceptable stress levels.

Conclusions:

• The developed AI model demonstrates performance comparable to gastrointestinal surgeons.
• AI-driven visualization of dissectable layers may reduce intraoperative recognition errors.
• The AI system shows potential for enhancing safety and efficiency in gastrointestinal surgery.