GL-FusionNet: Fusing global and local features to classify deep and superficial partial thickness burn.

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare
• Burn wound assessment

Background:

• Burns are common and painful injuries requiring accurate depth classification.
• Distinguishing between superficial partial thickness and deep partial thickness burns is challenging for inexperienced clinicians.
• Automated and accurate burn depth classification is needed to improve patient care.

Purpose of the Study:

• To develop an automated deep learning model for accurate burn depth classification.
• To improve the efficiency and accuracy of burn wound assessment in clinical settings.
• To differentiate between superficial partial thickness and deep partial thickness burns.

Main Methods:

• Utilized U-Net for burn wound segmentation, achieving high Dice and IoU scores.
• Proposed GL-FusionNet, a novel classification model fusing local features (ResNet50) and global features (ResNet101).
• Employed clinically collected, physician-labeled burn images for training and validation.

Main Results:

• U-Net segmentation achieved Dice score of 85.352 and IoU score of 83.916.
• The GL-FusionNet model demonstrated superior performance in burn depth classification.
• Achieved high classification metrics: 93.523% accuracy, 93.67% recall, 93.51% precision, and 93.513% F1-score.

Conclusions:

• The proposed deep learning approach, GL-FusionNet, offers an accurate and automated solution for burn depth classification.
• This method significantly enhances the efficiency of initial burn diagnosis and nursing care.
• The model shows potential for widespread clinical adoption to aid medical staff in burn assessment.