Automated Non-Invasive Burn Diagnostic System for Healthcare using Artificial Intelligence: AMBUSH-AI.
Authors
Affiliations (7)
Affiliations (7)
- Department of Surgery, School of Medicine, University of Pittsburgh, Pittsburgh, PA, 15219.
- McGowan Institute for Regenerative Medicine, University of Pittsburgh, Pittsburgh, PA, 15219.
- Department of Computer Science, Purdue University, West Lafayette, IN, 47907.
- School of Industrial Engineering, Purdue University, West Lafayette, IN 47907, USA.
- Division of Plastic Surgery, School of Medicine, Indiana University, Indianapolis, IN 46202.
- Department of Pathology and Laboratory Medicine, School of Medicine, Indiana University, Indianapolis, IN 46202.
- Department of Plastic Surgery, School of Medicine, University of Pittsburgh, Pittsburgh, PA, 15219.
Abstract
Develop technology to predict burn wound depth using a combination of FDA approved ultrasound modalities and interpretation of these images using artificial intelligence (AI). Physical examination by burn surgeons is the diagnostic gold standard to determine the need for burn surgery. Distinguishing between deep partial and third degree burns to determine the need for surgery is the ultimate diagnostic challenge. Reported accuracy for this process is 76% for burn experts and 50% for non-experts. A pig burn model (n=12) was used to develop the initial AI framework, which was subsequently tested in a nonrandomized prospective study of thermal burn human subjects (n=30). Images from Tissue Doppler Elastography Imaging (TDI), to measure tissue stiffness, Harmonic B-mode ultrasound, to identify anatomic landmarks, and digital photographs were collected. Biopsies were obtained from 5 subjects who went to the OR for debridement as ground truth for AI image interpretation. The AI model analyzed both TDI and B-mode images to predict burn depth. AI accuracy and explainability in predicting burn depth were the main outcomes. The AI algorithm identified third degree burns in pigs with 100% accuracy. For human subjects the mean age was 47.6± 17.6 years old and TBSA is 7.7±8.5%. The AI method achieved a 95% accuracy in identifying 3rd degree burns in humans. These results indicate that the strategy to use AI interpretation of B-mode ultrasound and TDI images to increase diagnostic accuracy in predicting burn depth is feasible.