Whale Optimization Algorithm with a Hybrid Relation Vector Machine: A Highly Robust Respiratory Rate Prediction Model...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Photoplethysmography (PPG) signals offer convenient dynamic monitoring of respiration rate (RR) compared to impedance spirometry.
• Accurate RR estimation from low-quality PPG signals, common in intensive care, remains a significant challenge.

Purpose of the Study:

• To develop a robust machine learning model for real-time RR estimation from PPG signals, specifically addressing low signal quality.
• To improve RR prediction accuracy by integrating signal quality metrics into the model.

Main Methods:

• A hybrid relation vector machine (HRVM) model was developed, optimized using the whale optimization algorithm (WOA).
• The model incorporated signal quality factors to enhance the robustness of RR estimation from PPG.
• Performance was validated using the BIDMC dataset, comparing PPG-derived RR with impedance spirometry.

Main Results:

• The proposed model achieved Mean Absolute Error (MAE) of 0.71 breaths/min and Root Mean Square Error (RMSE) of 0.99 breaths/min on the training set.
• On the test set, MAE was 1.24 breaths/min and RMSE was 1.79 breaths/min.
• Incorporating signal quality metrics reduced MAE and RMSE by up to 1.28 and 1.67 breaths/min (training) and 0.62 and 0.65 breaths/min (test) compared to models without these factors.

Conclusions:

• The WOA-HRVM model effectively estimates respiration rate from PPG signals, demonstrating significant accuracy improvements when considering signal quality.
• The method shows promise for reliable RR monitoring in challenging clinical scenarios with weak PPG signals.
• This approach offers a valuable tool for dynamic respiratory monitoring, particularly in intensive care settings.