Related Experiment Video
Updated: Jan 15, 2026
06:55
Protocol to Create Chronic Wounds in Diabetic Mice
Published on: September 25, 2019
21.6K
Machine learning model to classify chronic leg wounds and identify pyoderma gangrenosum
Dorothee A Busch1,2, Mats L Richter3,4, Jens Hüsers5
1Health Informatics Research Group, Osnabrück University of Applied Sciences, Faculty of Business Management and Social Sciences, Osnabrück, Germany d.busch@hs-osnabrueck.de.
BMJ Health & Care Informatics
|October 10, 2025
Summary
A new machine learning model accurately identifies pyoderma gangrenosum (PG) wounds, differentiating them from common diabetic foot ulcers (DFU) and venous leg ulcers (VLU). This AI tool aids in diagnosing this rare inflammatory skin disease.
Area of Science:
- Medical imaging analysis
- Artificial intelligence in dermatology
- Wound care diagnostics
Background:
- Chronic wounds pose significant economic and personal burdens.
- Accurate diagnosis is crucial for effective wound treatment.
- Pyoderma gangrenosum (PG) is a rare inflammatory skin disease often misdiagnosed.
Purpose of the Study:
- Develop a machine learning model to differentiate PG from other chronic wound types.
- Focus on chronic leg and foot wounds to address diagnostic challenges.
- Improve diagnostic accuracy for rare inflammatory skin conditions.
Main Methods:
- Utilized 3674 wound photographs from specialized centers.
- Included common leg/foot ulcers and pyoderma gangrenosum.
- Employed a ConvNeXt 'B' convolutional neural network classifier, pre-trained and fine-tuned.
Main Results:
- Achieved high accuracy in multiclass wound classification (90% unbalanced, 87% balanced).
- Demonstrated high sensitivity for PG (94%) and other common ulcers like DFU (97%) and VLU (92%).
- Observed higher misclassification rates for mixed and arterial leg ulcers.
Conclusions:
- The AI model effectively differentiates PG from common leg and foot ulcers.
- High accuracy was noted for diabetic foot ulcers and venous leg ulcers.
- Provides a foundation for a diagnostic support system for wound classification.
