[June 20th] The 26th Future of Surgery: AI-Powered Skin Ulcer Assessment – A First Step Towards Real-Time Wound Assessment
AI translatedAbout this session
Friday, June 20, 2025 18:00-20:00
keynote speech
"Minimally invasive reconstruction of intractable ulcers using capillary perforator flap"
Isao Mitsushima, Professor Emeritus, University of Tokyo
Invited lecture
"AI-based skin ulcer assessment – A first step towards real-time wound assessment"
Kazufumi Tate, Associate Professor, Department of Plastic Surgery, Faculty of Medicine, Tohoku Medical and Pharmaceutical University
Traditionally, the evaluation of skin ulcers has relied on visual inspection by physicians, but this approach is subjective and lacks reproducibility. Therefore, we are attempting AI-based image analysis of ulcers. Using our own software, we annotated ulcer images into six classes and performed semantic segmentation. While automatic ulcer recognition was good, accuracy was low for classes with ambiguous color. To compensate for this, we implemented a supplementary annotation mechanism using RGB values and a linear classification formula, and are currently using this in conjunction with AI judgment to improve accuracy. In the future, we aim to develop an AI camera that can evaluate ulcers in real time and track changes over time.
June 20th, Fri, 18:00-20:00(JST)
Keynote
Capillary Perforator Flap for Refractory Ulcers
- A Minimally Invasive Approach to Complex Wound Reconstruction -
Dr. Isao Koshima
Professor Emeritus, The University of Tokyo
Invited Lecture
Using AI to Evaluate Skin Ulcers:
A Step Toward Real-Time Wound Assessment
Dr. Kazufumi Tachi
Associate Professor, Division of Plastic and
Reconstructive Surgery,
Tohoku Medical and Pharmaceutical University
Skin ulcers have traditionally been assessed visually by physicians, but this approach is subjective and lacks reproducibility. To address this issue, we have been developing an AI system for automatic ulcer evaluation using image analysis. We annotated ulcer images into six classes using a custom-built software and trained deep learning models for semantic segmentation. While the models performed well in general, accuracy dropped in classes with ambiguous color features. To complement this, we implemented a secondary annotation system based on RGB values and linear classification formulas. Our current model combines AI-based and rule-based assessments to improve accuracy. Ultimately, our goal is to develop a real-time AI camera system that can evaluate and track ulcer progression over time.
|
skin ulcer treatment
AI diagnosis
Real-time evaluation
Predictive medicine
Negative pressure-assisted high-speed vascular anastomosis device
SkinUlcerTreatment
AIDiagnosis
Real-timeEvaluation
PredictiveMedicine
Negative-PressureAssistedAnastomoticDevice
|
Program
Koshima Isao
The Next Generation of Super Microsurgery International Forum, The Next-Generation of Super Mirosurgery International Forum
The University of Tokyo,
Tachi Kazufumi
Tohoku Medical and Pharmaceutical University,
Skin ulcers have traditionally been assessed visually by physicians, but this approach is subjective and lacks reproducibility. To address this issue, we have been developing an AI system for automatic ulcer evaluation using image analysis. We annotated ulcer images into six classes using a custom-built software and trained deep learning models for semantic segmentation. While the models performed well in general, accuracy dropped in classes with ambiguous color features. To complement this, we implemented a secondary annotation system based on RGB values and linear classification formulas. Our current model combines AI-based and rule-based assessments to improve accuracy. Ultimately, our goal is to develop a real-time AI camera system that can evaluate and track ulcer progression over time.