A grid matrix-based Raman spectroscopic method to characterize different cell milieu in biopsied axillary sentinel lymph nodes of breast cancer patients

Abstract

Raman spectroscopy which is based upon inelastic scattering of photons has a potential to emerge as a noninvasive bedside in vivo or ex vivo molecular diagnostic tool. There is a need to improve the sensitivity and predictability of Raman spectroscopy. We developed a grid matrix-based tissue mapping protocol to acquire cellular-specific spectra that also involved digital microscopy for localizing malignant and lymphocytic cells in sentinel lymph node biopsy sample. Biosignals acquired from specific cellular milieu were subjected to an advanced supervised analytical method, i.e., cross-correlation and peak-to-peak ratio in addition to PCA and PC-LDA. We observed decreased spectral intensity as well as shift in the spectral peaks of amides and lipid bands in the completely metastatic (cancer cells) lymph nodes with high cellular density. Spectral library of normal lymphocytes and metastatic cancer cells created using the cellular specific mapping technique can be utilized to create an automated smart diagnostic tool for bench side screening of sampled lymph nodes. Spectral library of normal lymphocytes and metastatic cancer cells created using the cellular specific mapping technique can be utilized to develop an automated smart diagnostic tool for bench side screening of sampled lymph nodes supported by ongoing global research in developing better technology and signal and big data processing algorithms.

Appendix 1

PCA/PC-LDA: PCA (principal component analysis) is a known data reduction technique where huge spectral data are decomposed into small independent variables known as “factors,” and contributions of these factors are called “scores”. In short, PCA involves a mathematical procedure that transforms a number of (possibly) correlated variables into a (smaller) number of uncorrelated variables called principal components. The first principal component accounts for as much of the variability in the data as possible, and each succeeding component accounts for as much of the remaining variability (lesser than the preceding principal component) as possible. Spectral data analysis is carried out over entire region and on the high frequency region following with plotting of factor 1 versus 2 in this technique.

Correlation and peak-to-peak ratio (PPR) analysis—Raman spectra signals

To analyze unknown/unbiased Raman spectra signals, basic concept of correlation rather than the complex signal discrimination techniques is used here. Correlation is a measure of the similarity between two signals, and when both the signals are same, it is called autocorrelation. In signal processing, cross-correlation is a measure of similarity of two waveforms as a function of a time-lag applied to one of them. It is commonly used for searching a long signal for a shorter, known feature. It also has applications in pattern recognition, single particle analysis, electron tomographic averaging, cryptanalysis, and neurophysiology.

Data Observation: Correlation—Any correlation coefficient value >0.9995 is considered to be considerably high, and the signals have high similarity. Any correlation coefficient value <0.9960 ± 0.0005 is considered to be quite different, and the signals have distinct features. Even though this may be minute difference for general signals, it is quite high variation for biomedical signals, which is also seen from the peak ratios. De-noised signals (using MATLAB command, cmd denoise) for all tissues were also analyzed for correlation coefficient value, and it gives 99% or more of the same correlation value, and in few cases, it gives a little higher (+0.0001) correlation coefficient value. It never gives lesser correlation (as compared to the noisy original signal) as the signal is de-noised with some minute peaks which may be important medically, flattened out.

Peak ratio—Peak ratio is taken on de-noised signals as the “ratio” will not change much even if we average out the signal. The intensity value changes by de-noising, but it changes almost equally at all points. The wave number range (cm⁻¹ as generated from the Raman spectra measurement device) from which specific peaks (the highest intensity peak in these range is chosen to take the ratio) are chosen is – Peak1 −804.029 to 911.677; Peak2 −922.363 to 1028.45; Peak3 −1233.01 to 1334.98; Peak4 −1398.86 to 1498.62; Peak5 −1606.83 to 1703.83; Peak6 −2814.44 to 2895.14. This is based on human observation of signals that major peaks occur at those particular locations consistently in all signals, with varying intensities. We observed same results with peak ratio analysis as obtained by cross-correlation.

Appendix

Correlation value for all tissues :-

In MATLAB (.mat) files

Peak ratio code for tissue 30 :- (Similar for all other tissues)