Affective Computing Based on Morphological Features of Photoplethysmography for Patients with Hypertension

Area of Science:

• Cardiology
• Psychophysiology
• Biomedical Engineering

Background:

• Emotions significantly impact hypertension risk and prognosis.
• Photoplethysmography (PPG) signals offer potential for physiological monitoring.
• Affective computing (AC) can analyze physiological data for emotional state detection.

Purpose of the Study:

• To discriminate emotional states in hypertension patients using PPG waveform indices and AC.
• To evaluate the effectiveness of AC in classifying emotional states based on PPG signals.
• To explore feature selection methods for optimizing AC performance.

Main Methods:

• Forty-three essential hypertension patients underwent PPG signal acquisition under various emotional conditions.
• Five key PPG waveform indices were extracted and analyzed.
• A support vector machine classifier was employed, with resubstitution and cross-validation for performance assessment.
• Feature selection using full search and genetic algorithms (GA) was performed.

Main Results:

• AC achieved 100% accuracy in distinguishing emotional states from baseline using resubstitution.
• Six-fold cross-validation demonstrated high accuracy with 10 waveform features.
• GA feature selection improved classification accuracy to 78.97% (2-class), 74.22% (3-class), and 67.35% (4-class).

Conclusions:

• The proposed AC method effectively categorizes emotional states in hypertension patients using five PPG waveform indices.
• This approach demonstrates high accuracy and potential for non-invasive emotional state monitoring in clinical settings.
• The study highlights the utility of specific PPG indices and AC for understanding the interplay between emotion and hypertension.