The transformation of normal cervix to cervicitis as well as to cervical cancer is accompanied with biochemical alterations at cellular level. Laser induced fluorescence (LIF) can reflect those changes either as variations in the fluorescence intensity or as shift in the fluorescence maxima of bio fluorophores present in tissues. The curve resolved fluorescence investigation of tissues under 325 nm excitation provides collagen, bound nicotinamide adenine dinucleotide (NADH) and free NADH as the discrimination factors between normal, cervicitis and cervical cancer. Even though the fluorescence emission intensity derived from collagen fiber is comparable in both normal and cervicitis, a considerable reduction was observed for the cervical cancer tissues compared to the former. Fluorescence corresponding to bound NADH is found to be reduced during the progression from normal to cervicitis and to cervical cancer, whereas the free NADH shows an opposite trend. The principal component analysis (PCA) was performed to obtain classification of spectral data from different categories on a reduced dimensional space. Furthermore, to test the usefulness of the recorded fluorescence spectra in discriminating the malignant and non-malignant (cervicitis and normal) samples, a supervised machine learning model based on support vector machine (SVM) was built using the PCA-reduced data. The proposed SVM model was able to detect the malignant samples with a sensitivity of 94.19% and specificity of 96.51%. Moreover, the Raman spectral data from the corresponding tissue sites corroborate well with the observations derived from the fluorescence measurement. The results obtained in the present pilot study strongly suggests the potential of LIF technique combined with multivariate data analysis tool for the diagnosis of cervicitis and cervical malignancy.
All Science Journal Classification (ASJC) codes
- Electronic, Optical and Magnetic Materials
- Atomic and Molecular Physics, and Optics