Baseline removal from near infrared spectroscopy measurements for lactate concentration estimation

Abstract

A smoothness priors approach for removing the baseline from near infrared (NIR) spectroscopy measurements is presented. Typically in NIR spectroscopy applications the baseline is removed by subtracting the mean or a linear trend from the data. The smoothness priors method is basically a time-varying high-pass filter with adjustable cut-off frequency and it has been used in various applications where trend removal is essential. The proposed method is applied to NIR spectroscopy measurements of 28 different concentrations of lactate (concentrations varied between 0-40 mmol) in 2% intralipid solution. The measurements are analyzed by using principal component regression (PCR) method. The effect of the baseline removal is evaluated by comparing the results obtained with the smoothness priors method to those obtained with traditional linear baseline removal.