ABC-YOLO: Automated skin burn depth classification using YOLO architectures

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Accurate skin burn depth classification is crucial for effective treatment and patient recovery.
• Current diagnostic methods can be subjective, leading to potential delays in care.
• Automated classification systems can improve diagnostic speed and objectivity.

Purpose of the Study:

• To compare the performance of various YOLO-based deep learning models for automated skin burn depth classification.
• To identify the most effective YOLO architecture for this specific medical imaging task.
• To assess the potential of deep learning as a clinical decision support tool for burn management.

Main Methods:

• A multi-source dataset was compiled, including hospital records and public image repositories.
• Images were meticulously labeled into four burn degrees by expert general surgeons.
• Segmentation-based YOLOv8 and YOLOv11 models of varying sizes were trained and evaluated.
• Data augmentation and preprocessing techniques were employed to optimize model performance.

Main Results:

• The YOLOv11x-seg model significantly outperformed other tested architectures.
• YOLOv11x-seg achieved an F1-Score of 0.87 and a mAP@0.5 of 0.91.
• Statistical analysis confirmed the superior performance and significance of the YOLOv11x-seg model.

Conclusions:

• The YOLOv11x-seg architecture demonstrates high accuracy and potential for automated skin burn classification.
• This deep learning model can serve as a valuable, rapid, and objective decision support tool in clinical practice.
• The study contributes to advancing burn diagnosis through the integration of state-of-the-art AI in medical image analysis.