Machine Learning-Based Classification of Syncope and Presyncope Using Heart Rate Variability

Syncope is also known as fainting, while presyncope is a sensation of fainting that does not result in loss of consciousness. The standard diagnostic test for these conditions is the Head-Up Tilt Test (HUTT), which is uncomfortable, somewhat invasive and requires special equipment. This work investigated a possible alternative test using heart rate variability (HRV) analysis. Wavelet analysis was applied to the heart rate time series consisting of inter-beat intervals, and summarised using the mean, variance, skewness and kurtosis. Significant differences were found between syncope and presyncope groups. These significant results persisted over a range of wavelet scales. A further analysis using machine learning classification provided a significant discrimination between syncope and pre-syncope with an accuracy of 80%. The results have implications for understanding these conditions and the possible links between disease and the autonomic system. In addition, such analysis may provide an alternative diagnostic tool, avoiding the expensive and invasive HUTT and enhancing patient comfort.