Ensemble Extreme Learning Machine Method for Hemoglobin Estimation Based on PhotoPlethysmoGraphic Signals.

Area of Science:

• Biomedical Engineering
• Medical Diagnostics
• Machine Learning Applications

Background:

• Accurate hemoglobin (Hb) concentration measurement is crucial for clinical health screening and blood transfusions.
• Existing non-invasive Hb detection methods require improvements in accuracy and stability for clinical applicability.

Purpose of the Study:

• To develop and evaluate a novel non-invasive method for detecting hemoglobin (Hb) concentration.
• To enhance the accuracy and stability of non-invasive Hb detection using advanced machine learning techniques.

Main Methods:

• Established a mathematical model based on the Beer-Lambert law for non-invasive Hb detection.
• Utilized eight-wavelength PhotoPlethysmoGraphic (PPG) signals, followed by denoising and feature extraction.
• Employed Recursive Feature Elimination (RFE) for feature selection and an Ensemble Extreme Learning Machine (EELM) for regression modeling.

Main Results:

• The proposed EELM method achieved a Root Mean Square Error (RMSE) of 1.72 g/dL.
• A strong Pearson Correlation Coefficient (PCC) of 0.76 (p < 0.01) was observed between predicted and reference Hb values.
• The method was validated on a dataset of 249 clinical cases (199 training, 50 testing).

Conclusions:

• The developed non-invasive Hb detection method shows significant potential for clinical applications.
• The EELM regression model based on PPG signals offers a stable and accurate alternative to invasive Hb measurements.