Two-step regression analysis approach to assess burn wound severity using spatial frequency domain imaging.

Area of Science:

• Biomedical Optics
• Wound Healing Research
• Medical Imaging Analysis

Background:

• Prompt burn wound care is critical for successful recovery.
• Spatial Frequency Domain Imaging (SFDI) shows potential for predicting burn healing outcomes.
• Calibrated reflectance images (Rd) from SFDI offer a non-invasive approach to assess burn severity.

Purpose of the Study:

• To develop predictive models for burn severity using SFDI.
• To estimate collagen denaturation depth (CDD) on day 1 postburn using SFDI Rd.
• To predict day 28 healing outcomes based on day 1 CDD.

Main Methods:

• Utilized a porcine burn model with graded severity and varying contact durations.
• Collected SFDI and color images on multiple days post-injury (0-28 days).
• Performed histological analysis for CDD assessment and clinical evaluation for healing outcomes.
• Applied a two-step regression framework: linear regression for CDD prediction and logistic regression for healing outcome prediction.

Main Results:

• SFDI Rd accurately predicted day-1 CDD with a root mean square error of 105 μm and R² of 0.71.
• The logistic model predicted day-28 healing outcomes with an ROC-AUC of 0.88.
• CDD was confirmed as an early indicator of burn severity and healing potential.

Conclusions:

• A two-step framework using SFDI Rd enables early prediction of burn severity and healing potential.
• This approach allows for non-invasive assessment as early as day 1 postburn.
• Findings may simplify future SFDI hardware design by clarifying contrast sources.