Aim: To evaluate the accuracy of a combined artificial intelligence and multispectral imaging device (SpectralAI) in predicting burn wound healing within three weeks, compared to clinical observation and Specialist Burn Consultant opinion.
Methods: A prospective study was conducted in adult patients with partial-thickness burns. Patients were scanned using the SpectralAI Device at the time of their initial presentation following burn injury. Predictions were compared with actual healing times in patients that were clinically predicated to heal or the decision for operative intervention in patients clinically predicted not to heal. The primary outcome was the accuracy of the device in correctly predicting healing within three weeks.
Results: 23 patients with 61 burn sites were included. The SpectralAI device showed a sensitivity of 70.5% and a specificity of 100%. Further patient recruitment and data collection is ongoing. Comparative analysis between SpectralAI predictions and expert assessment revealed areas for improvement in device reliability.
Conclusion: SpectralAI shows promise in burn depth assessment and healing prediction. Further multi-centre validation with larger patient cohorts is needed to refine the accuracy and reliability of the device. The integration of this technology with clinical expertise could enhance burn care decision-making.
Keywords: burn depth, artificial-intelligence, multispectral, SpectralAI