Pulmonary Hypertension Detection Non-Invasively at Point-of-Care Using a Machine-Learned Algorithm.

Area of Science:

• Medical research
• Cardiovascular diagnostics
• Artificial intelligence in medicine

Background:

• Pulmonary hypertension (PH) diagnosis is challenging due to its complex nature and overlapping symptoms with other cardiovascular diseases.
• Non-invasive diagnostic methods are crucial for early detection and management of PH.

Purpose of the Study:

• To develop and validate a supervised machine learning model for non-invasive diagnosis of pulmonary hypertension.
• To assess the model's performance using orthogonal voltage gradient and photoplethysmographic signals.

Main Methods:

• A supervised machine learning model was trained using 3298 features extracted from non-invasive orthogonal voltage gradient and photoplethysmographic signals.
• Model performance was evaluated using sensitivity, specificity, and Area Under the Receiver Operator Characteristic Curve (AUC-ROC).
• Feature importance analysis was conducted to identify key contributors to the model's predictive power.

Main Results:

• The developed model achieved a sensitivity of 87% and a specificity of 83%, with an AUC-ROC of 0.93.
• Consistent performance was observed across different genders, age groups, and classes of pulmonary hypertension.
• Analysis identified changes in conduction, repolarization, and respiration metrics as significant predictors.

Conclusions:

• The machine learning model demonstrates significant potential for accurate, non-invasive diagnosis of pulmonary hypertension.
• This approach could facilitate early detection and timely intervention when integrated into point-of-care diagnostic systems.