Spectral region optimization for Raman-based optical diagnosis of inflammatory lesions

FT-Raman Spectroscopy was applied to identify biochemical alterations existing between inflammatory fibrous hyperplasia (IFH) and normal tissues of buccal mucosa. One important implication of this study is related to the cancer lesion border. In fact, the cancerous normal border line is characterized by the presence of inflammation and its correct discrimination would increase the accuracy in delimiting the lesion frontier. Seventy spectra of IFH from 14 patients were compared to 30 spectra of normal tissue from 6 patients. The statistical analysis was performed with Principal Components Analysis and Soft Independent Modeling Class Analogy methodologies. After studying several spectral ranges it was concluded that the best discrimination capability (sensibility of 95% and specificity of 100%) was found using the 530 to 580 cm$^{-1}$ wavenumbers. The bands in this region are related to vibrational modes of Collagen aminoacids Cistine, Cysteine, and Proline and their relevant contribution to the classification probably relies on the extracellular matrix degeneration process occurring in the inflammatory tissues.

💡 Research Summary

This study investigates the potential of Fourier‑transform Raman (FT‑Raman) spectroscopy to differentiate inflammatory fibrous hyperplasia (IFH) from normal buccal mucosa, with the broader aim of improving the delineation of cancer lesion borders where inflammation frequently co‑exists. Seventy Raman spectra were collected from 14 patients diagnosed with IFH (approximately five spectra per patient) and thirty spectra from six healthy volunteers (five spectra per subject). All measurements were performed using a 1064 nm Nd:YAG laser FT‑Raman system, covering the 400–1800 cm⁻¹ range with a spectral resolution of 4 cm⁻¹. Each spectrum resulted from three 10‑second accumulations to enhance signal‑to‑noise ratio.

Data preprocessing involved baseline correction, total‑area normalization, and second‑derivative calculation to sharpen peak features and suppress fluorescence background. Outliers were removed based on Mahalanobis distance thresholds. The preprocessed dataset was subjected to unsupervised Principal Component Analysis (PCA) for dimensionality reduction and visual inspection of clustering, followed by supervised Soft Independent Modeling of Class Analogy (SIMCA) to build class‑specific models for IFH and normal tissue. Model performance was evaluated via leave‑one‑out cross‑validation.

A systematic “window‑search” was conducted by training separate PCA‑SIMCA models on successive 20 cm⁻¹ spectral windows. The window spanning 530–580 cm⁻¹ yielded the highest discriminative power, achieving 95 % sensitivity (19 of 20 IFH spectra correctly identified) and 100 % specificity (all normal spectra correctly classified). This narrow region encompasses vibrational modes primarily associated with the side‑chain sulfhydryl groups of cysteine and the ring breathing modes of proline, both integral components of collagen’s triple‑helix structure. In IFH samples, these peaks exhibited reduced intensity and slight wavenumber shifts, reflecting extracellular matrix (ECM) remodeling and collagen degradation driven by inflammatory processes such as matrix metalloproteinase activity.

PCA score plots using the first three principal components (explaining ~78 % of total variance) clearly separated the two groups, with IFH spectra clustering on the negative side of PC2 and normal spectra on the positive side. SIMCA’s confidence ellipses confirmed that all normal spectra fell well within the normal class region, while only one IFH spectrum lay near the decision boundary, accounting for the observed 95 % sensitivity.

The clinical implication is significant: during oral cancer surgery, the tumor‑normal interface often harbors an inflammatory component that can obscure histopathological assessment. A rapid, non‑destructive optical technique capable of detecting such inflammation could guide surgeons in real‑time, reducing both positive margins (residual tumor) and unnecessary removal of healthy tissue. By focusing on the 530–580 cm⁻¹ band, a compact Raman probe could be engineered for intra‑operative use, delivering immediate feedback without the need for tissue excision or staining.

Limitations of the current work include the modest sample size and the single‑center recruitment, which may limit generalizability across diverse patient populations and lesion subtypes. Moreover, Raman spectra are sensitive to experimental variables such as temperature, hydration, and laser power; thus, standardization of acquisition protocols is essential for clinical translation. Future research should expand the cohort, incorporate other inflammatory and pre‑malignant oral lesions, and explore advanced multivariate classifiers (e.g., support vector machines, deep learning) that can integrate information from multiple spectral regions simultaneously. Combining Raman spectroscopy with complementary optical modalities such as optical coherence tomography (OCT) could also enable three‑dimensional mapping of lesion boundaries.

In conclusion, the study demonstrates that a narrowly defined Raman spectral window (530–580 cm⁻¹) provides a robust biochemical fingerprint for distinguishing IFH from normal buccal mucosa, achieving near‑perfect specificity and high sensitivity. This finding underscores the promise of Raman‑based optical diagnostics as a real‑time adjunct for accurately defining cancer margins in the oral cavity, potentially improving surgical outcomes and patient quality of life.