Raja Roy1, Shatakshi Srivastava1, Vivek Gupta2, Ashish Tiwari2, Anand N. Srivastava3, Abhinav A. Sonkar2
1Centre of Biomedical Magneetic Resonance, Lucknow, Uttar Pradesh, India; 2Departments of General Surgery, Chattrapati Shahuji Maharaj Medical University, Lucknow, Uttar Pradesh, India; 3Departments of Pathology, Chattrapati Shahuji Maharaj Medical University, Lucknow, Uttar Pradesh, India
In the present work, HR-MAS NMR spectroscopic studies have been performed on human oral SCC tumor tissues, its neighboring margins and bed tissues (n=159), obtained from 36 patients (n=27 training set; n=9 unknown test set), for the identification of metabolic fingerprints. The proton NMR spectra were then subjected to PCA, OSC-filtered PCA and PLS-DA multivariate analysis. The training data-set (n=120 tissue specimens; 27 patients) of PLS-DA model allowed correct classification of malignant tissues from benign samples with >98% specificity and sensitivity. The class membership of unknown tissue specimens (n=39) obtained from nine patients were classified with 97.4% diagnostic accuracy.