A Method to Create a Universal Calibration Dataset for Raman Reconstruction Based on Wiener Estimation

Raman spectroscopy has been intensively explored in biomedical applications. However, Raman data acquisition is generally slow due to inherently weak Raman signals. Narrow-band Raman measurements can compensate for the weak Raman signal by performing integration in the wavenumber dimension, in which the Raman spectrum with high spectral resolution needs to be reconstructed from the narrow-band measurements. However, this method is limited in the requirement of a calibration dataset, in which the calibration samples have to be similar to test samples in Raman features. Therefore, a new calibration dataset is often needed for every type of samples. We propose a method to create a universal calibration dataset for Raman reconstruction to overcome this limitation. In our method, the Raman spectra measured from each basic biochemical component in samples instead of from actual samples are used in the calibration dataset. In this study, the proposed method was tested on 27 liquid phantoms and 56 cell measurements. The results demonstrated the excellent performance of Raman reconstruction using the universal calibration dataset compared to that using the traditional calibration dataset and the measured Raman spectra.