RRI-based machine learning system for early warning and emergency alerts of refractory vasovagal syncope
X Y Yang, R H Zhang, C ChengAbstract
Background
Vasovagal Syncope (VVS) is the most common form of neurocardiogenic syncope, and traditionalmethods rely on the head upright tilt test (HUTT), which has limited diagnostic value due to its low sensitivity.Sudden syncope and falls have resulted in high rates of disability and psychological burdens, constituting asignifificant public health concern. Artifificial intelligence (AI) technology have been applied in disease diagnosisand risk prediction.Therefore, developing intelligent diagnosis and risk prediction for VVS is crucial to preventadverse syncope events.
Purpose
Developing intelligent diagnosis and risk prediction for early warning and emergency alerts of refractoryVVS to prevent adverse syncope events.
Methods
We developed a explainable deep learning model named VVS-AI, which can diagnose VVS and provideearly warnings based solely on the RRI obtained during the HUTT.
Results
In our internal testing, the model successfully predicted a syncope event 61 RR intervals in advance witha specifificity of 1.00 and a sensitivity of 0.944. In the external test set, our system issued early warning signalsfor all HUTT-positive patients, achieving a sensitivity of 84.6% and a specifificity of 100%. The system effificiencyexceeded 90%.
Conclusion
AI technology can provide early warnings for VVS, and its low computational cost makes it suitablefor integration into portable health monitoring devices for daily use and emergency intervention of VVS eventsfigure--1.pngfigure 2positive test patients