Development of new classification models based on Raman spectroscopy and MALDI spectrometry as novel tools for liver cancer diagnostic

Hepatocellular carcinoma is the most common type of primary liver tumor. Usually occurring along with liver fibrosis or cirrhosis, it is often hard to diagnose. Therefore, novel optical diagnostic tools are currently explored that are able to detect various biomarkers in cytological and histological samples. In recent years Raman Spectroscopy and MALDI imaging have been applied for cancer diagnostics. The objective of the present thesis was to investigate the applicability of those techniques as diagnostic tools for liver cancer detection. In the first study cells from liver cancer cell lines were analyzed by Raman imaging. A support vector machine classification model resulted in a prediction accuracy of 93% for cells. By applying hierarchical cluster analysis to each single cell, different cellular compartments such as nucleus, lipid droplets and cytoplasm were differentiated. Using only spectral information of lipids, the prediction accuracy of classification model improved to 96% in comparison to 91% for nuclei, 87% for cytoplasm and 93% for the complete cell information. To investigate the diagnostic capacities of Raman spectroscopy on tissue level patient samples were analyzed. We were able to detect lipids, proteins, collagen and cholesterol ester as separate components within the Raman maps. A random forest classification model allowed us to predict tissue regions with sensitivity of 76% and specificity of 93%. As a second diagnostic technique for analyses of different proteins within tissue regions MALDI was applied. Four proteins 6274, 6647, 6222 and 6853 m/z with significantly higher expression profile in the HCC tissue regions in comparison with non-tumorous liver tissue were identified. The developed classification model allowed prediction of HCC with sensitivity and specificity of 90%. In conclusion, the obtained results showed that Raman and MALDI IMS imaging techniques can successfully detect and predict liver cancer on cellular and tissue level.