Burn depth assessment
Key Points
- Do not rely solely on clinical assessment for indeterminate-depth burns -- accuracy is only 64-76% even among experienced surgeons [3][4]
- Laser Doppler imaging remains the most studied adjunct and can predict healing potential approximately 2 days ahead of clinical judgment [3]
- Infrared thermography correlates well with LDI and offers a simpler, more cost-effective alternative, though its low sensitivity (44.5%) limits standalone use [1]
- Machine learning models show promise with depth classification accuracy above 83%, but lack prospective clinical validation [2][5]
- Reassess wound depth serially -- initial clinical impressions frequently change as the wound declares itself over the first 48-72 hours [4][6]
Overview
Burn depth assessment determines whether a wound will heal spontaneously within 21 days or requires excision and grafting. For indeterminate-depth burns, clinical assessment by experienced burn surgeons achieves an accuracy of only 64% to 76% [3]. This error rate has direct surgical consequences: unnecessary operations on wounds that would have healed, or delayed grafting on wounds that will not. Adjunct technologies can improve accuracy, especially in the 48-72 hour window when clinical signs are most ambiguous.
Pathophysiology
The 21-day healing threshold drives surgical decision-making. Burns that will not re-epithelialize by day 21 benefit from early excision and grafting, while those that will heal spontaneously are best managed conservatively. The challenge is that indeterminate-depth burns -- partial-thickness injuries whose ultimate depth is unclear in the first days -- may evolve as the wound declares itself over 48-72 hours [4][6].
Assessment
Clinical examination
Clinical assessment by experienced burn surgeons achieves 64-76% accuracy for predicting which wounds will heal within 21 days [3]. This has been the standard of care for decades but carries a clinically significant misclassification rate for indeterminate-depth burns [4].
Laser Doppler imaging
Laser Doppler imaging (LDI) estimates wound perfusion, with lower perfusion correlating with deeper injury. Some investigators report 96-100% accuracy and the ability to predict healing potential approximately 2 days ahead of clinical judgment, though others question its clinical applicability [3]. LDI remains the most studied adjunct technology for burn depth assessment.
Infrared thermography
Infrared thermography showed statistically significant correlations with LDI in 4 of 4 clinical studies, with a mean specificity of 98.8% for predicting healing within 15 days, though sensitivity was lower at 44.5% [1]. It offers a simpler, more cost-effective alternative to LDI but cannot be used as a standalone tool given its low sensitivity.
Machine learning and artificial intelligence
Machine learning models for burn depth classification achieved accuracies exceeding 83%, while TBSA estimation models performed comparably to or better than paper-based methods [2]. A meta-analysis of 12 studies found burn depth accuracy of 68.9-95.4% with a pooled sensitivity of 90.8% and specificity of 84.4% [5]. These models remain in early development with limited prospective clinical validation.
Enzymatic debridement as diagnostic adjunct
Enzymatic debridement with anacaulase-bcdb offers an alternative approach to depth assessment by enabling selective eschar removal with dermal preservation, addressing the dual challenge of accurate depth determination and appropriate surgical intervention [8].
Controversies and Evidence Gaps
Most adjunct technology studies involve small, single-center cohorts, and heterogeneity in study design limits meta-analytic synthesis. Infrared thermography demonstrates high specificity but low sensitivity, meaning it is better at confirming deep wounds than detecting all wounds that will fail to heal. Machine learning models remain in early development with limited validation in prospective clinical settings. Key controversies include the cost and availability of LDI, the clinical applicability of thermography given its low sensitivity, and the readiness of AI models for prospective clinical use.
References
[1] Dang et al. (2021). Use of Infrared Thermography for Assessment of Burn Depth and Healing Potential: A Systematic Review. PMID: 34120173 [2] Huang et al. (2021). A systematic review of machine learning and automation in burn wound evaluation: A promising but developing frontier. PMID: 34419331 [3] Jaskille et al. (2010). Critical review of burn depth assessment techniques: part II. Review of laser doppler technology. PMID: 20061851 [4] Heimbach et al. (1992). Burn depth: a review. PMID: 1290249 [5] Taib et al. (2023). Artificial intelligence in the management and treatment of burns: A systematic review and meta-analyses. PMID: 36571960 [6] Singer et al. (2017). Burn Wound Healing and Tissue Engineering. PMID: 28328668 [7] Zuo et al. (2017). Important Developments in Burn Care. PMID: 27294857 [8] Shoham et al. (2023). Anacaulase-bcdb for the treatment of severe thermal burns. PMID: 37833828