Research on a Burn Severity Detection Method Based on Hyperspectral Imaging.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Dermatology

Background:

• Accurate burn wound detection is critical for managing infection risk and hypertrophic scarring.
• Current clinical judgment for burn severity has limited accuracy (65-70%).
• Non-invasive methods are needed for efficient and precise burn assessment.

Purpose of the Study:

• To develop a non-invasive, efficient method for burn severity assessment using hyperspectral imaging (HSI).
• To introduce a novel deep learning model, MBNet, for improved burn depth diagnosis.
• To evaluate MBNet's performance against existing machine learning algorithms.

Main Methods:

• Utilized hyperspectral imaging (HSI) for contactless spectral detection of burn-affected skin.
• Developed a burn severity detection classification network (MBNet) based on the Mamba model.
• Employed a bidirectional scanning strategy within MBNet to capture long-term spectral dependencies.

Main Results:

• MBNet effectively distinguishes subtle spectral differences across various burn conditions.
• The proposed MBNet model demonstrated significantly higher accuracy than seven other machine learning algorithms on a custom dataset.
• HSI shows potential for precise monitoring of structural changes in burn-affected tissue.

Conclusions:

• MBNet offers a reliable and efficient approach for clinical burn severity assessment.
• Hyperspectral imaging, coupled with advanced deep learning, holds significant promise for improving burn diagnosis.
• Further research is needed to facilitate the widespread clinical adoption of HSI for burn management.