When analyte concentrations are predicted from spectral data, the mean prediction error decreases as the signal-to-noise of the spectrum increases. However, medical situations often place an upper limit on the amount of time that a clinical measurement can take. The time constraint, in turn, limits the signal-to-noise ratio that can be achieved. Therefore, when we design calibrations for analyte prediction, it is important to study how the prediction accuracy changes as a function of collection time. Our preliminary investigations, using our data on blood serum, have indicated that we can decrease the measurement time for unknown samples to as little as 30 or 60 seconds while preserving a clinical level of prediction accuracy.
Showing the most recent 10 out of 178 publications