Diagnostic potential of near-infrared Raman spectroscopy in the stomach: differentiating dysplasia from normal tissue

Abstract

Raman spectroscopy is a molecular vibrational spectroscopic technique that is capable of optically probing the biomolecular changes associated with diseased transformation. The purpose of this study was to explore near-infrared (NIR) Raman spectroscopy for identifying dysplasia from normal gastric mucosa tissue. A rapid-acquisition dispersive-type NIR Raman system was utilised for tissue Raman spectroscopic measurements at 785 nm laser excitation. A total of 76 gastric tissue samples obtained from 44 patients who underwent endoscopy investigation or gastrectomy operation were used in this study. The histopathological examinations showed that 55 tissue specimens were normal and 21 were dysplasia. Both the empirical approach and multivariate statistical techniques, including principal components analysis (PCA), and linear discriminant analysis (LDA), together with the leave-one-sample-out cross-validation method, were employed to develop effective diagnostic algorithms for classification of Raman spectra between normal and dysplastic gastric tissues. High-quality Raman spectra in the range of 800-1800 cm(-1) can be acquired from gastric tissue within 5 s. There are specific spectral differences in Raman spectra between normal and dysplasia tissue, particularly in the spectral ranges of 1200-1500 cm(-1) and 1600-1800 cm(-1), which contained signals related to amide III and amide I of proteins, CH(3)CH(2) twisting of proteins/nucleic acids, and the C=C stretching mode of phospholipids, respectively. The empirical diagnostic algorithm based on the ratio of the Raman peak intensity at 875 cm(-1) to the peak intensity at 1450 cm(-1) gave the diagnostic sensitivity of 85.7% and specificity of 80.0%, whereas the diagnostic algorithms based on PCA-LDA yielded the diagnostic sensitivity of 95.2% and specificity 90.9% for separating dysplasia from normal gastric tissue. Receiver operating characteristic (ROC) curves further confirmed that the most effective diagnostic algorithm can be derived from the PCA-LDA technique. Therefore, NIR Raman spectroscopy in conjunction with multivariate statistical technique has potential for rapid diagnosis of dysplasia in the stomach based on the optical evaluation of spectral features of biomolecules.

Figure 1

Photomicrographs of the haematoxylin and eosin (H&E)-stained tissue sections of gastric tissues (A) normal and (B) dysplasia (high-grade dysplasia of the antrum). Scale bar: 100 μm.

Figure 2

Mean normalised gastric Raman spectra (solid line) ±1 s.d. (grey area) obtained from a normal tissue (A) and a dysplasia tissue (B) by multiple measurements (n=5) at various locations for each sample. Each spectrum was normalised to the integrated area under the curve to correct for variations in absolute spectral intensity. All spectra were acquired in 5 s with 785-nm excitation and corrected for spectral response of the system.

Figure 3

Mean normalised gastric Raman spectra ±1 s.d. (shaded area) from (A) normal tissues (n=55) and (B) dysplasia tissues (n=21), illustrating the intensity variations in major Raman peaks of 20–30% for normal tissues whereas of 30–60% for dysplasia tissues.

Figure 4

(A) Comparison of the mean normalised Raman spectra of normal (n=55) and dysplasia (n=21) tissues. (B) Difference spectrum calculated from the mean Raman spectra of normal and dysplasia tissue (i.e., the mean normalised Raman spectrum of dysplasia tissue minus the mean normalised Raman spectrum of normal tissue).

Figure 5

Scatter plot of the intensity ratio of Raman signals at 875 and 1450 cm⁻¹, as measured for each sample and classified according to the histological results. The decision line (I₈₇₅/I₁₄₅₀=0.717) separates dysplasia tissue from normal tissue with a sensitivity of 85.7% (18/21) and specificity of 80.0% (44/55).

Figure 6

The first four diagnostically significant principal components (PCs) accounting for about 78.5% of the total variance calculated from Raman spectra (PC1 – 42.6%, PC2 – 25.4%, PC4 – 7.9%, and PC5 – 2.6%), revealing the diagnostically significant spectral features for tissue classification.

Figure 7

Scatter plots of the diagnostically significantly principal component (PC) scores for normal and dysplastic gastric tissue derived from Raman spectra, (A) PC1 vs PC2; (B) PC1 vs PC4; (C) PC1 vs PC5; (D) PC2 vs PC4; (E) PC2 vs PC5; and (F) PC4 vs PC5. The dotted lines (PC2=1.46 PC1+1.34; PC4=−1.32 PC1+0.94; PC5=−2.16 PC1−0.89; PC4=1.74 PC2+0.12; PC5=0.84 PC2−0.381; and PC5=−2.05 PC4−0.29) as diagnostic algorithms classify dysplasia from normal with sensitivity of 90.5% (19/21), 76.2% (16/21), 71.4% (15/21), 81.0% (17/21), 71.4% (15/21), and 71.4% (15/21); specificity of 90.9% (50/55), 80.0% (44/55), 83.6% (46/55), 80.0% (44/55), 72.7% (40/55), and 72.7% (40/55), respectively. Circle (○): normal; Triangle (▴): dysplasia.

Figure 8

Scatter plot of the linear discriminant scores for the normal and dysplasia categories using the PCA-LDA technique together with leave-one-spectrum-out, cross-validation method. The separate line yields a diagnostic sensitivity of 95.2% (20/21) and specificity of 90.9% (50/55) for differentiation between normal and dysplasia tissue. LDA, linear discriminant analysis; PCA, principal components analysis.

Figure 9

Comparison of receiver operating characteristic (ROC) curves of discrimination results for Raman spectra utilising the PCA-LDA-based spectral classification with leave-one-spectrum-out, cross-validation method and the empirical approach using Raman intensity ratio of I₈₇₅/I₁₄₅₀. The integration areas under the ROC curves are 0.98 and 0.88 for PCA-LDA-based diagnostic algorithm and intensity ratio algorithm, respectively, demonstrating the efficacy of PCA-LDA algorithms for tissue classification. LDA, linear discriminant analysis; PCA, principal components analysis.