Range-independent Background Subtraction Algorithm for Recovery of Raman Spectra of Biological Tissue
by Shovan K. Majumder
Laser Biomedical Applications and Instrumentation Division, Raja Ramanna Centre for Advanced Technology, Indore, India
Dr Shovan K. Majumder from the Laser Biomedical Applications and Instrumentation Division, Raja Ramanna Centre for Advanced Technology, Indore, India discusses the development of a range-independent background subtraction algorithm for recovery of Raman spectra of biological tissue  .
What was the purpose of your research?
The purpose of our research was to develop an automated background subtraction algorithm whose output is immune to the choice of the fitting range specified by its user, and enables faithful recovery of artifact-free Raman signatures, characteristic of biological tissue, by subtracting the order of magnitude stronger background from the experimentally measured raw tissue Raman spectra.
What were the key results from your research?
A range-independent algorithm (RIA) capable of rapid retrieval of Raman signatures by automated background subtraction from the measured raw tissue Raman spectra was developed. A comparison with two most widely used polynomial-based algorithms (ModPoly and I-ModPoly) using mathematically simulated Raman spectra having different curvatures and signal-to-baseline ratios as well as experimentally measured Raman spectra from various biological samples were found to yield consistently range-independent and artifact-free Raman signal with zero baseline. Further, the RIA was found to fulfill all the requirements of an effective background subtraction method that is desired for successful use of Raman spectroscopy for real-time, noninvasive, automated diagnosis of various cancers in a clinical situation.
What does this actually mean?
Successful use of in vivo Raman spectroscopy for biological applications requires engaging an appropriate method for extracting rather weak Raman signals from the broad, orders of magnitude stronger, background believed to be arising from fluorescence and/or the scattering tail of the laser line. Although the polynomial-based, particularly the modified polynomial- based, algorithms are known to be superior to most of the remaining background subtraction algorithms in their ability for simple and effective background reduction and are extensively used, they have one major shortcoming. They are sensitive to the choice of the spectral region to be used in the fit. Thus selection of different start and stop wavenumbers leads to Raman spectra of significantly different line shapes and intensities which turn out to be a nuisance in the subsequent quantitative analysis of the spectral data. This effect is the most prominent if any of these wavenumbers happens to be in a Raman active region particularly at the position of a Raman peak or on the leading or trailing slope of a Raman band. Moreover, the polynomial- based algorithms, in general, also suffer from their dependence on the order of the polynomials used in the iterative fit. The background subtraction algorithm developed by us overcomes all these shortcomings of the polynomial-based methods while facilitating faithful representation of the extracted Raman spectra from the raw spectra measured from biological samples. The algorithm was found to provide good range independence and retrieval of all the Raman peaks by efficiently subtracting the associated intense background.
Figure 1: (a) Raw synthetic Raman spectrum and the Raman signals recovered from it using (b) the RIA, (c) the ModPoly, and (d) the I-ModPoly for three different spectral ranges: (i) range-1 corresponding to 800-1800 cm-1, (magenta) (ii) range-2 corresponding to 980-1580 cm-1 (red), and (iii) range-3 corresponding to 1150-1750 cm-1 (green).
1. Hemant Krishna, Shovan K. Majumder and Pradeep K. Gupta, Journal of Raman Spectroscopy, 43(12), 1884–1894 (2012).