The Role of Artificial Intelligence in Diagnosis and Prognosis of Kawasaki Disease: A Cross-Sectional Study in Mashhad, Iran

Introduction: Kawasaki Disease (KD) is a leading cause of pediatric acquired heart disease globally, with diagnostic challenges and unpredictable prognosis. Early detection and risk stratification are critical to prevent complications. Recent advancements in Artificial Intelligence (AI) offer promising tools for enhancing diagnostic accuracy and prognostic assessment. This study aims to evaluate the application of AI algorithms in diagnosing KD and predicting its outcomes in a tertiary hospital setting. Materials and Methods: This retrospective cross-sectional study included ۲۵۰ children diagnosed with KD at Akbar Hospital, Mashhad, Iran, between ۲۰۱۸ and ۲۰۲۳. Demographic, clinical, laboratory, and echocardiographic data were collected. AI models, including machine learning classifiers such as Random Forest, Support Vector Machine, and Neural Networks, were trained to recognize patterns associated with KD diagnosis and to predict cardiac complications such as coronary artery dilation. Data was split into training and testing sets (۷۰/۳۰), with model performance evaluated using accuracy, sensitivity, specificity, and area under the ROC curve. Results: The AI models demonstrated high diagnostic accuracy, with the Neural Network achieving an accuracy of ۹۲%, sensitivity of ۹۰%, and specificity of ۹۳%. For prognosis prediction, models successfully identified children at higher risk of coronary artery involvement with an accuracy of ۸۸%. Key clinical and laboratory variables influencing AI predictions included elevated C-reactive protein, erythrocyte sedimentation rate, and serum sodium levels. The models outperformed traditional diagnostic criteria in sensitivity and allowed for early risk stratification. Conclusion: AI algorithms show significant potential in improving the accuracy and timeliness of Kawasaki Disease diagnosis and prognosis prediction. Implementing AI-based tools in clinical practice could enhance early detection, enable personalized treatment plans, and reduce cardiovascular complications. Further prospective studies are warranted to validate these findings and facilitate integration into routine pediatric care.