Multivariate analysis of the Raman and NIR fluorescence spectra for optical biopsy of skin cancer
During the translation of the optical biopsy method to the clinical settings there is a decrease in the useful signal level for in vivo measurements. To determine the possibility to reach the required level of diagnostic accuracy in clinical application the in vivo pilot clinical study of the malignant and benign tumors was carried out in Samara oncological hospital. The human tissue spectra simultaneously including Raman and fluorescence signals were registered using by the portable spectroscopic equipment with 785 nm excitation laser wavelength in the 800-920 nm near infrared region. To improve the analysis efficiency of the spectral signal with high noise level the multivariate method for analysis of the Raman bands with broadband fluorescence in the full spectral regions 800-920 nm. The informative spectral bands for different neoplasm type that correspond to the both fluorescence spectral shape and Raman bands were identify by the analysis of the variable importance in projection analysis. The spectral data were analyzed using PLS-DA method. The results of the skin neoplasm differentiating were presented with boxplot diagrams of the PLS coefficients, that predict of the diagnose for each neoplasm, and receiving operating curve (ROC). The proposed analysis methods allowed to achieve the differentiating accuracy more than 89%. To estimate of the efficiency of the Raman and autofluorescence detection the comparative analysis of the optical differentiating accuracy and the provisional diagnosis of the oncologist was carried out. The results show that proposed method of the tissue optical control have a significant clinical potential due to specific advantages of the spectroscopic methods that realized in the single portable equipment.
File with abstract
Samara National Research University
File with report