A prospective cohort study testing whether oral cavity mucosal tissue variability influences elastic scattering spectroscopy (ESS) measures.

OBJECTIVE

Elastic Scattering Spectroscopy (ESS) has been shown to accurately discriminate between benign and malignant tissue in the oral cavity, which makes it a promising technology to assist with intraoperative oral cancer margin guidance. However, some spectroscopy-based technologies are influenced by locoregional histopathological differences, which can influence diagnostic accuracy and data acquisition procedures. The objective of this study was to assess the extent to which ESS spectral readings of mucosal tissue varied between different anatomic regions of the oral cavity in a cohort of healthy adults.

METHODS

The study included 30 participants with no known history of smoking, vaping, or other substance use-related carcinogenic exposures, with normal oral mucosa, and without history of head or neck cancer or other inflammatory conditions. ESS measurements were obtained at predefined anatomic sites within the oral cavity, and spectra were grouped into five classes based on histopathologic similarities of tissue architecture (Class 1: soft palate, mucosal lip, buccal mucosa; Class 2: retromolar trigone, hard palate, gingiva; Class 3: dorsal tongue; Class 4: lateral tongue; Class 5: floor of mouth, ventral tongue). Random forests were used to build predictive models and Leave-One-Patient-Out (LOPO) cross-validation used to obtain performance statistics. Statistical analyses included sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), and overall model accuracy.

RESULTS

Overall analysis yielded sensitivity = 52.2%, specificity = 88.4%, and accuracy = 0.52. The overall mean weighted F1 score was 0.53. Using a one-vs-all approach of the five anatomic sites, sensitivity = 43.4%-62.1%, specificity = 85%-91%, PPV = 29.2%-61.5%, and NPV = 78.3%-95.5%.

CONCLUSION

Discriminatory accuracy between pairs of oral cavity subsites was moderate, but still significantly above random chance (0.20 for five classes). This supports the contention that the success of ESS in identifying malignancy is robust against locoregional variations in "normal" tissues within the oral cavity. These data further support the application of ESS as a valuable adjunct tool for oral cancer margin guidance by demonstrating that ESS is tissue agnostic in the oral cavity.