Determination of glycated hemoglobin using near-infrared spectroscopy combined with equidistant combination partial least squares

• We developed a simple method to determine glycated hemoglobin in blood.
• The method was based on near-infrared spectroscopy with stable wavelength selection capability.
• The sensitivity and specificity values for diabetes were above 93.5%.

A novel near-infrared-spectroscopy-based quantification method for glycated hemoglobin (HbA1c), a major clinical diagnosis indicator of diabetes, was developed on the basis of simultaneous determination of hemoglobin (Hb) and absolute HbA1c content (Hb
• HbA1c) in human hemolysate samples. Equidistant combination partial least squares (EC-PLS) method was proposed to perform wavelengths selection. Competitive adaptive reweighted sampling PLS (CARS-PLS) and Monte Carlo uninformative variable elimination PLS (MC-UVE-PLS) methods were also conducted for comparison. A randomness and stability dependent rigorous process of calibration, prediction, and validation was performed to produce objective and stable models. The search range covered the unsaturated region (780–1880 nm, 2090–2330 nm). For Hb and Hb
• HbA1c, only 6 and 14 wavelengths were selected with EC-PLS, 23 and 30 wavelengths were selected with CARS-PLS, and 100 and 120 wavelengths were selected with MC-UVE-PLS, respectively.The predicted values of relative percentage HbA1c were calculated from the predicted Hb and Hb
• HbA1c values. The sensitivity and specificity for diabetes were 93.5% and 97.1% with EC-PLS, 91.3% and 94.1% with CARS-PLS, and 89.1% and 76.5% with MC-UVE-PLS, respectively. In three methods, EC-PLS not only employed the least wavelengths but also produced the best quantification accuracy for HbA1c. EC-PLS also achieved the highest classification accuracy for negative and positive samples for diabetes.The results confirm the feasibility of HbA1c quantification based on the simultaneous analysis of Hb and Hb
• HbA1c with NIR spectroscopy. This technique is rapid and simple when compared with conventional methods, and is a promising tool for screening diabetes in large populations.