Automated anxiety detection using probabilistic binary pattern with ECG signals.

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Cardiology

Background:

• Anxiety disorders are prevalent, necessitating early detection for effective management.
• Anxiety impacts physiological systems, including the brain and heart.
• Electrocardiography (ECG) signals offer a window into cardiac responses to anxiety.

Purpose of the Study:

• To develop an efficient and accurate handcrafted feature engineering model for automated anxiety detection.
• To leverage ECG signals for classifying anxiety levels using machine learning.

Main Methods:

• Utilized open-access ECG data from 19 subjects exposed to anxiety-inducing videos.
• Employed a novel probabilistic binary pattern (PBP) feature extraction method combined with tunable q-factor wavelet transform.
• Applied dimensionality reduction techniques (Neighborhood Component Analysis, Chi2) and machine learning classifiers (k-NN, SVM) with a greedy algorithm for optimal model selection.

Main Results:

• Achieved classification accuracies exceeding 98.5% across all analyzed ECG segment lengths (4, 5, and 6 seconds).
• Ablation studies validated the superior performance of PBP-based feature engineering compared to traditional methods like Local Binary Patterns.

Conclusions:

• The developed PBP-based feature engineering model demonstrates high feasibility and accuracy for anxiety classification using ECG signals.
• This automated approach holds promise for objective and early detection of anxiety disorders.