Multispectral Imaging for Hemoglobin Estimation by PCA.

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Physiology

Background:

• Assessing tissue blood perfusion is crucial for monitoring health and superficial lesions.
• Current methods for blood flow monitoring often lack non-invasiveness, quantitation, or broad applicability.
• There is a need for accessible, large-area, non-invasive blood flow assessment techniques.

Purpose of the Study:

• To develop and validate a non-invasive method for quantifying tissular blood perfusion using multispectral imaging.
• To estimate changes in deoxygenated hemoglobin concentration as an indicator of blood flow.
• To provide a quantitative measure of blood perfusion in the visible and near-infrared (VIS-NIR) spectrum.

Main Methods:

• Utilizing multispectral imaging in the VIS-NIR range to capture tissue optical properties.
• Applying principal component analysis (PCA) for data dimensionality reduction.
• Employing a linear regression model to correlate spectral data with deoxygenated hemoglobin changes.

Main Results:

• The proposed method achieved an average prediction similarity of 91.53% against photoplethysmography measurements.
• The model demonstrated a mean adjusted R-squared value of 0.8104, indicating strong predictive performance.
• Validation was performed using in-vivo data from a vascular occlusion protocol.

Conclusions:

• Multispectral imaging combined with PCA and linear regression offers a viable non-invasive approach for quantifying blood perfusion.
• The method accurately estimates changes in deoxygenated hemoglobin, correlating well with established physiological measurements.
• This technique holds potential for improved health assessment and lesion monitoring through accessible blood flow quantification.