Using Modeling and Machine Learning to Characterize POTS

Postural Orthostatic Tachycardia Syndrome (POTS) causes autonomic nervous system dysfunction resulting in excessive increase in heart rate in response to a postural change. Symptoms, including dizziness upon standing, fainting, and widespread pain, overlap with a variety of other autonomic conditions, making it difficult to diagnose. POTS is often observed after an immunological stressor like disease or vaccination, but not much is known about its physiological causes. The head-up tilt test (HUT) is traditionally used to diagnose POTS. While this test is effective, it requires special equipment, takes at least 20 minutes, and can result in fainting. The Valsalva Maneuver (VM) is an alternative test used to diagnose autonomic function in less than a minute with a breath hold instead of a postural change, making it low risk and highly accessible. The objective of this study is to use mathematical modeling, parameter estimation, and machine learning to test if the VM can provide a reliable method to diagnose POTS. We obtained data from ~700 patients who exhibited side effects of the HPV vaccine. Our patient-specific model uses blood pressure data to predict heart rate, estimating identifiable parameters that minimize the least squares error between the model predictions and data. Using machine learning, we will analyze uncorrelated and optimized model parameters and quantities extracted from data to (i) determine if VM data and patient-specific model parameters can be used to identify POTS patients, and (ii) identify what markers best characterize POTS patients.