An Image Based Object Recognition System for Wound Detection and Classification of Diabetic Foot and Venous Leg...

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare
• Chronic wound management

Background:

• Venous leg ulcers and diabetic foot ulcers are prevalent chronic wounds with increasing incidence.
• These conditions strain healthcare resources due to their complexity and management needs.
• Accurate and efficient wound assessment is crucial for effective treatment and patient outcomes.

Purpose of the Study:

• To evaluate the performance of YoloV5 detection and classification algorithms for venous leg ulcers and diabetic foot ulcers.
• To assess the utility of AI in identifying and categorizing common chronic wound types from images.
• To explore the potential for automated wound data entry in patient records.

Main Methods:

• Application of YoloV5 family pre-trained models to a dataset of 885 wound images.
• Specific focus on the YoloV5m6 model for detection and classification tasks.
• Quantitative evaluation using precision, recall, and mean Average Precision (mAP) metrics.

Main Results:

• The YoloV5m6 model achieved a high precision of 0.942 and a recall of 0.837.
• The mAP_0.5:0.95 score was 0.642, comparable to existing literature.
• Precision and recall values significantly surpassed those previously reported in similar studies.

Conclusions:

• The YoloV5m6 model demonstrates strong performance in detecting and classifying chronic leg and foot ulcers.
• AI-driven wound analysis offers a promising avenue for semi-automated patient record integration.
• Future work includes developing a clinician dashboard to validate AI predictions and build trust.