Multimodal diagnostic segregation of oral leukoplakia and cancer

Oral leukoplakia (OLK) is the most common pre malignant disorder (PMD) with highest malignant potentiality. It is clinically highly correlated with oral squamous cell carcinoma (OSCC). Painful biopsy is the gold standard till date for diagnosis of these diseases. Again for specific grading of such pathological states and mitigation of inter and intra observer variability and subjective disease classification, alternative of molecular biomarkers for diseases differentiation as well as non-invasive modalities are yet to be explored. In this study role of morphometric, intensity and textural features extracted from liquid based exfoliative cytology (LBEC) and intensity and textural features extracted from ex vivo optical coherence tomography (OCT) images and spectral features from difference between mean spectra (DBMS) were evaluated for disease classification using variants of support vector machine (SVM) classifiers. Result showed that at 10 fold cross-validation OLK and OSCC could be differentiated using cellular features of LBEC data at 100% sensitivity and specificity. Spectral biomarkers were also extracted efficiently which could classify the diseases with 81.3% sensitivity and 91.3% specificity depicting role of chemical molecules responsible in pathological alteration. Considering the advantage of each modality, it can be concluded that these features can utilized as disease differentiation markers and have useful clinical implication for diagnosis of OLK and OSCC towards mitigation of inter- as well as intra observer variability faced during routine histopathological diagnostic procedure.