A measurement system must be calibrated before it can quantify the concentration of analytes in a blood sample. State-of-the-art calibration methods employ spectroscopy, which can yield noisy results and have trouble obtaining data from a dynamically changing sample. Traditional attempts to solve these problems with spectroscopy require training data from multiple non-continuous spectroscopy “gold standard” measurements. This results in an over-trained model that cannot be applied generally between samples. In a clinical setting, these traditional methods require drawing small quantities of blood with each measurement, which often compromises the measurement system calibration due to sample perturbation and sample-to-sample variability. Thus, there is a need for a non-invasive calibration method that requires minimal “gold standard” measurements, allows for the continuous measurement of analytes within an individual, and does not compromise the sample.