AI Model for Identifying Loose Connective Tissue in Surgery
Author Information
Author(s): Kumazu Yuta, Kobayashi Nao, Senya Seigo, Negishi Yuya, Kinoshita Kazuya, Fukui Yudai, Mita Kazuhito, Osaragi Tomohiko, Misumi Toshihiro, Shinohara Hisashi
Primary Institution: Anaut Inc.
Hypothesis
Can an AI model effectively recognize and display loose connective tissue as a dissectable layer during gastrointestinal surgery?
Conclusion
The AI model demonstrated performance comparable to that of experienced gastrointestinal surgeons in recognizing loose connective tissue.
Supporting Evidence
- The AI detected at least 75% of loose connective tissue in 91.8% of the images evaluated.
- Surgeons judged that in 99% of the AI videos, visualization was improved.
- False positives were found in 52.6% of the images, but most were not significant enough to affect surgical judgment.
Takeaway
The AI can help doctors see important tissue layers during surgery, making it easier and safer to operate.
Methodology
The AI model was trained on over 30,000 annotated images from surgical videos and evaluated by 10 external surgeons.
Potential Biases
The evaluation excluded images with obvious false positives, which may not reflect real-world performance.
Limitations
The AI model was unapproved for clinical use and tested in a simulated environment, not during actual surgeries.
Participant Demographics
The study involved gastrointestinal surgeons from multiple hospitals in Japan.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website