Terahertz spectral imaging for early assessment of frostbite injuries using the double Debye model and supervised machine learning

Early assessment of the severity of frostbite injuries is critical for guiding clinical management and improving patient outcomes; however, tissue damage evolves dynamically and is difficult to predict during the acute phase. Terahertz time-domain spectroscopy (THz-TDS) has previously demonstrated high accuracy in the early triage of thermal burn injuries. In this study, we evaluate the potential of the THz-TDS modality using a portable handheld scanner to assess the depth of frostbite wounds. A standardized in vivo porcine frostbite model was employed, and injury classification was performed using a support vector machine algorithm. We demonstrate that the area under the receiver operating characteristic (ROC) curves was 0.94, 0.85, and 0.87 for healthy tissue, partial-thickness frostbite, and full-thickness injuries. In addition, we explored the use of the double Debye dielectric relaxation model of tissue to reduce the data dimensionality. We observed significant statistical differences between the double Debye parameters of the three groups. These results demonstrate the potential of THz-TDS imaging for early, non-destructive assessment of the depth of frostbite injuries and suggest its potential utility in improving clinical decision-making and surgical outcomes.