Optimized Signal Quality Assessment for Photoplethysmogram Signals Using Feature Selection.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Wearable Technology

Background:

• Remote patient monitoring relies on wearable healthcare devices.
• Accurate interpretation of Photoplethysmographic (PPG) signals is crucial for cardiovascular health assessment.
• Signal quality assessment (SQA) is essential for reliable PPG data.

Purpose of the Study:

• To develop and optimize an SQA approach for PPG signals from wearable devices.
• To enhance the accuracy of remote patient monitoring through improved signal quality assessment.

Main Methods:

• An ensemble-based feature selection scheme was employed.
• Optimized feature subsets were identified to differentiate clean and artifact-corrupted PPG segments.
• Classification models were trained and validated on diverse PPG datasets.

Main Results:

• The proposed feature selection approach achieved high discriminatory power.
• Accuracy, sensitivity, and specificity rates exceeded 0.93, 0.89, and 0.97, respectively, on both dependent and independent test datasets.
• Performance was superior to contemporary methods across various conditions, including atrial fibrillation and normal sinus rhythm.

Conclusions:

• The developed SQA method demonstrates robustness against dynamic variations in PPG signals.
• The approach is reliable for various wearable devices and long-term monitoring.
• This enhances confidence in using wearable PPG devices for remote cardiovascular monitoring.