Near infrared spectroscopy for hemoglobin quantification in a single drop of blood with improved pretreatment and wavelength selection

Abstract

Clinical diagnostics and patient monitoring depend heavily on hemoglobin (HGB) quantification; however, traditional techniques are frequently intrusive, time-consuming, and require comparatively large blood volumes. By investigating the possibility of using Near Infrared (NIR) spectroscopy to predict hemoglobin concentration from a single drop of blood, this study addresses the demand for a quick and minimally invasive alternative. The primary goal is to increase prediction accuracy by using optimized wavelength selection and spectral preprocessing. A total of 400 blood samples, including those from healthy individuals and patients receiving medical treatment with hemoglobin values ranging from 2.5 to 20 g/dL, were collected from AL QODS Medical Analysis Laboratory. To improve spectral clarity, spectral preprocessing methods such as the Savitzky–Golay Method (SGM), Standard Normal Variate (SNV), and Multiplicative Scatter Correction (MSC) were applied. An optimal subset of wavelengths was then selected using the Interval Partial Least Squares (IPLS) regression technique. Compared to conventional Partial Least Squares (PLS) using the entire spectrum, the results demonstrate that IPLS significantly improves model performance. A strong correlation between laboratory data and a selection of 39 spectral bands (2238.758–2307.405 nm) was also observed. This study demonstrates how advanced preprocessing and wavelength selection can be combined to quantify hemoglobin from a single blood drop accurately. The results highlight the potential of NIR spectroscopy as a fast, reliable, and easily accessible clinical monitoring tool, particularly for the diagnosis and monitoring of anemia and other blood disorders.