An empirical investigation of deviations from the Beer–Lambert law in optical estimation of lactate

Abstract The linear relationship between optical absorbance and the concentration of analytes—as postulated by Beer-Lambert law—is one fundamental assumptions that much spectroscopy literature is explicitly or implicitly based upon. common use regression models such as principal component partial least squares exemplifies how linearity assumption upheld in practical applications. However, also establishes deviations from law can be expected when (a) light source far monochromatic, (b) concentrations analytes are very high (c) medium highly scattering. lack a quantitative understanding nonlinearities become predominant, along with mainstream nonlinear machine learning different fields, have given rise to methods random forests, support vector regression, neural networks spectroscopic This raises question that, small number samples variables many datasets, effects significant enough justify additional model complexity? In present study, we empirically investigate this relation lactate, an important biomarker. Particularly, analyze scattering matrices, three datasets were generated varying lactate phosphate buffer solution, human serum, sheep blood. Additionally, fourth dataset pertained invivo, transcutaneous spectra obtained healthy volunteers exercise study. Linear fitted each measures performance compared attest linearity. To isolate concentrations, solution was augmented six (100–600 mmol/L). Subsequently, partly overlapping extracted 0–11, 0–20 0–600 mmol/L. Similarly, dataset. analysis did not provide any evidence substantial due concentrations. results suggest may media, justifying complex, models.