Classification of orthostatic intolerance through data analytics.

Area of Science:

• Cardiovascular physiology
• Computational biology
• Clinical diagnostics

Background:

• Autonomic nervous system imbalance causes orthostatic intolerance (OI), leading to symptoms like dizziness and syncope.
• OI is challenging to diagnose due to unclear triggers and pathophysiology, resulting in classification variability.
• Accurate diagnosis and classification of OI subtypes are crucial for effective patient management.

Purpose of the Study:

• To apply machine learning for categorizing patients with orthostatic intolerance.
• To identify key markers from blood pressure and heart rate time-series data for classification.
• To distinguish between single and mixed pathophysiological conditions in OI patients.

Main Methods:

• Utilized random forest classification trees to identify predictive markers from time-series data.
• Employed K-means clustering to group identified markers.
• Analyzed clinical data from 186 subjects, including controls and patients with POTS, cardioinhibition, vasodepression, or mixed types.

Main Results:

• Achieved over 95% accuracy in categorizing patients with single OI conditions.
• Successfully subgrouped all patients with mixed cardioinhibitory and vasodepressor syncope.
• Clustering confirmed disease groups and revealed two distinct subgroups within control and mixed OI populations.

Conclusions:

• Machine learning effectively discovers structure in physiological time-series data for OI classification.
• The methodology aids in addressing diagnostic challenges and characterizing pathophysiological mechanisms of OI.
• This study represents a significant step toward using AI to support clinicians and researchers in OI diagnosis.