Automatic Wound Type Classification with Convolutional Neural Networks

Area of Science:

• Medical image analysis
• Artificial intelligence in healthcare
• Wound healing research

Background:

• Chronic wounds, such as diabetic foot and venous leg ulcers, require timely identification for effective healing.
• Accurate diagnosis of chronic wound types often necessitates specialized clinical expertise.
• Shortage of expert knowledge in certain healthcare settings can impede proper wound management.

Purpose of the Study:

• To evaluate the efficacy of a deep convolutional neural network (CNN) in classifying diabetic foot ulcers and venous leg ulcers using wound imagery.
• To assess the performance of the AI model in distinguishing between different types of chronic wounds.

Main Methods:

• A deep convolutional neural network (CNN) was developed and trained using 863 cropped images of chronic wounds.
• The model's classification performance was evaluated on a separate hold-out test set comprising 80 wound images.
• Performance metrics were analyzed for both cropped wound images and full wound images.

Main Results:

• The CNN model achieved an F1-score of 0.85 when classifying cropped wound images.
• The model demonstrated a slightly lower performance with an F1-score of 0.70 on full wound images.
• These results indicate a strong potential for AI in wound type classification.

Conclusions:

• The study demonstrates promising results for using deep learning models to classify chronic wounds, aiding clinical decision-making.
• The developed AI model shows potential for supporting clinicians in identifying diabetic foot and venous leg ulcers.
• Further research with expanded datasets and diverse wound types is recommended for clinical implementation.