DeepTriage: A Real-Time AI Decision Support System for Emergency Resource Allocation in Mass Casualty Incidents

Mass casualty incidents (MCIs) present significant challenges to emergency medical systems, frequently overwhelming available resources and necessitating complex triage decisions under severe time constraints. This paper introduces DeepTriage, a real-time artificial intelligence decision support system designed to optimize emergency resource allocation during MCIs. The system employs a hybrid neural network architecture that integrates convolutional, recurrent, and graph neural networks to process multimodal patient data and generate prioritization recommendations. The DeepTriage framework incorporates privacy-preserving mechanisms, dynamic resource optimization algorithms, and an adaptive transmission strategy for deployment in bandwidth-constrained environments. Performance evaluation conducted across three diverse datasets - TRAUMA-DB, MCI-SIM, and DISASTER-NET - demonstrates superior triage accuracy (92.6%) compared to traditional protocols (76.4-79.2%) and existing computational systems (84.5-87.3%). The system achieves significant improvements in decision speed (14.8s under benchmarked test conditions versus 187.3-245.8s for manual methods in similar scenarios) while maintaining a resource utilization efficiency of 0.87. DeepTriage exhibits robust performance across multiple incident types with minimal degradation under increasing patient loads. Implementation considerations address integration pathways with existing electronic health record systems, training requirements for medical personnel, and ethical frameworks governing algorithmic decision support in life-critical scenarios. The results indicate substantial potential for AI-driven systems to enhance emergency response capabilities during mass casualty incidents through improved triage accuracy, resource optimization, and decision consistency.