Burn characterization using object-oriented hyperspectral image classification.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Dermatology

Background:

• Accurate burn depth classification is crucial for effective treatment and surgical planning.
• Current methods for burn assessment can be subjective and lack detailed spatial information.

Purpose of the Study:

• To develop and validate a novel hyperspectral imaging (HSI) approach for objective burn depth classification.
• To generate detailed burn depth maps for improved characterization of burn wounds.

Main Methods:

• Acquisition of hyperspectral images from 14 burn patients using a pushbroom HSI system.
• Analysis of HSI data using an object-oriented classification (OOC) method, considering spectral, textural, and spatial attributes.
• Performance evaluation using metrics such as overall accuracy, precision, sensitivity, and specificity.

Main Results:

• The proposed HSI and OOC method achieved high performance in differentiating burn classes.
• Overall accuracy reached 95.99% ± 0.60%, with precision of 97.30% ± 2.46%, sensitivity of 97.23% ± 3.02%, and specificity of 98.02% ± 1.98%.

Conclusions:

• The object-based HSI classification approach effectively generates burn depth maps.
• This method offers a precise tool to assist surgeons in identifying and managing different depths of burn wounds.