Artificial Intelligence-assisted Raman Spectroscopy for Liver cancer diagnosis

Abstract

Hepatocellular carcinoma (HCC), the most common form of primary liver cancer, represents a global health challenge due to its complexity and the limitations of current diagnostic techniques. By combining Raman spectroscopy and Artificial Intelligence (AI), we have succeeded in classifying tumor cells. In fact, we have performed a first Raman spectral analysis based on the characterization and differentiation between uncultured primary human liver cells derived from resected HCC tumor tissue and the adjacent non-tumor counterpart. Biochemical analysis of the collected Raman spectra revealed that there is more DNA in the nuclei of the tumor cells than in non-tumor cells. We then develop three machine learning approaches, including multivariate models and neural networks, to rapidly automate the recognition and classification of the Raman spectra of both cells. To evaluate the performance of the developed AI models, we prepared and analyzed two additional cell samples with a ratio of 4:1 and 3:1 between tumor and non-tumor cells and compared the obtained results with the nominal percentages (accuracy of 80 and 60%, respectively). These results confirm that the models are able to make classifications at the level of a single spectrum, indicating the possibility of rapidly analysing and classifying a primary HCC cell.