Cluster analysis of lung adenocarcinoma cells using PCA and confocal Raman spectroscopy

This work analysis composition and structure of lung adenocarcinoma cells and normal cells based on confocal Raman spectroscopy. Firstly, confocal Raman spectra of 242 lung adenocarcinoma cells and 231 normal cells are obtained. In addition, Raman spectra are denoised and two evaluation criteria are used to evaluate the Raman spectra. Then, PCA is used to make principal component analysis of the spectrum and observe the process of spectral information changing with PCs. The best number of features is selected to prevent cross-sensitivity and obtain a better model. On this basis, a clustering model of PCA is constructed. The results show that accuracy of 84.57%, sensitivity of 88.02%, specificity of 80.95% and marshall correlation coefficient (MCC) of 69.21% are obtained. The results show that this method is effective in diagnosing cancer cells. Therefore, this model proves the potential of nondestructive detection of lung adenocarcinoma, and has a certain guiding role for the next classification and location. In addition, it can be easily applied in other fields in the future.