Improved Design of a Raman-based Optical Sensor for the Identification of Oral Cancer

Conclusion: In this thesis, the high potential of Raman spectroscopy (RS) to differentiate tumour from normal tissue was studied. Because of its capability to provide an objective "chemical fingerprint" of the examined region, RS can be used for the identification of molecular changes through specific spectral patterns in oral cancer tissue. I showed the potential of identifying oral tumour with high sensitivity and specificity in an ex vivo setting.

The hardware and software of the former set up was improved to reach high spectral resolution and intensity with lower excitation power (100 mW) and very short integration time (100 ms) giving rise to high diagnostic speed of the clinical Raman sensor, too. Hardware development was carried out through utilizing the Ventana spectrometer with twice higher spectral resolution than the QE65000 spectrometer (that was calculated theoretically and was confirmed by the experiment) and high sensitivity, wide numerical aperture hand-held Raman probe which has 2.5 times larger solid angle than the former Raman probe.

Development of software was done with the aim of having highly automated spectral data acquisition and analysis. Thus, a robust and user-friendly software, which can synchronize all process of shifting wavelength, spectra acquisition and spectral analysis for physicians was established which allows for a time-saving and cost-reducing application.

The suppression of the fluorescence background was done with the SERDS (shifted-excitation Raman difference spectroscopy) method followed by a polynomial fit. And, the reconstructed Raman spectrum of normal tissue represented intense peaks at 1258, 1308, 1458 and 1667 cm-1. In contrast, the Raman spectrum of oral tumour had lower peak intensities at 1291, 1342, 1458 and 1667cm-1. Hence, the Raman spectral difference between normal and tumour oral tissue of human was used for tumour diagnosis. There are obvious shifts from 1258cm-1 of normal tissue to 1291cm-1 peak of tumour tissue and from 1308cm-1 peak of normal tissue to 1342cm-1 peak of tumour tissue that were applied as criterion for differentiation. Besides, there is a clear intensity difference between normal and tumour tissue in 1458 cm-1 peak that was also used for discrimination. Accordingly, a statistical method will be implemented for the classification of the reconstructed Raman signal to convert the Raman spectrum into a diagnosis that would be carried out by my colleague in the other study.