Characterization of glycated hemoglobin based on Raman spectroscopy and artificial neural networks

Area of Science:

• Biomedical Spectroscopy
• Analytical Chemistry
• Diabetes Diagnostics

Background:

• Glycated hemoglobin (HbA1c) is the gold standard for diabetes diagnosis.
• Understanding HbA1c's spectral properties is crucial for diagnostic advancements.
• Raman spectroscopy offers a label-free method for molecular characterization.

Purpose of the Study:

• To analyze the Raman spectral characteristics of glycated hemoglobin (HbA1c).
• To develop a predictive model for quantifying HbA1c concentrations using Raman spectroscopy.
• To provide molecular assignments for observed Raman peaks in HbA1c.

Main Methods:

• Analysis of Raman spectra from commercial lyophilized HbA1c at various concentrations (4.76%, 9.09%, 100%).
• Identification and assignment of vibrational Raman peak positions.
• Development of a nonlinear regression model using a Feed-Forward Neural Network (FFNN).

Main Results:

• Identified characteristic Raman peaks for HbA1c powder at specific wavenumbers (e.g., 1578, 1436, 969 cm⁻¹).
• Achieved a Root Mean Square Error in Cross-Validation (RMSECV) of 0.08% ± 0.04% for HbA1c quantification.
• Provided detailed molecular assignments for the average spectra of lyophilized HbA1c.

Conclusions:

• Raman spectroscopy effectively characterizes HbA1c.
• The FFNN model demonstrates high accuracy in quantifying HbA1c concentrations.
• This approach holds promise for non-invasive diabetes diagnosis and monitoring.