Iterative refinement of a histologic algorithm for burn depth categorization based on 798 consecutive burn wound biopsies.

Document Type

Article

Publication Date

11-7-2023

Publication Title

Burns

Abstract

Introduction: Our group previously reported a burn biopsy algorithm (BBA-V1) for categorizing burn wound depth. Here, we sought to promulgate a newer, simpler version of the BBA (BBA-V2). Methods: Burn wounds undergoing excision underwent 4 mm biopsies procured every 25 cm2. Serial still photos were obtained at enrollment and at excision intraoperatively. Burn wounds assessed as likely to heal by 21 days were imaged within 72 h of injury and at 21 days. A sample of 798 burn wound biopsies were classified by both BBAV1 and BBAV2 algorithms. For nonoperative burn wounds, the proportion of healing versus nonhealing pixels at 21 days after injury were compared. Results: The 798 biopsies were classified by BBAV1 as 24% SPT, 47% DPT, 28% FT and by BBAV2 as 3% SPT, 67% DPT, and 30% FT (p < 0.0001). Overall, the proportion of biopsies whose wound reclassification changed from a nonoperative to operative pathway was 21% (95% CI: 18-24%). Nonoperative wounds judged at injury as being SPT contained 12.8 million pixels. Repeat 21-day imaging revealed 11.3 million healed pixels (accuracy = 89.6% (95% CI: 89.59-89.62)). Conclusions: BBA-V2 was associated with a significantly higher concordance with visual assessment for burn wounds clinically judged as deep partial and full thickness. Keywords: Algorithm; Artificial intelligence; Burn biopsy; Burn depth; Human.

First Page

23

PubMed ID

38040616

Volume

50

Issue

1

Publisher

Elsevier B.V.

ISBN

03054179

Link to Full Text

Share

COinS