Confocal Raman Sensing Based on a Support Vector Machine for Detecting Lung Adenocarcinoma Cells

This paper presents a label-free method for distinguishing lung adenocarcinoma cells from normal cells based on the confocal Raman spectra and a support vector machine. First, confocal Raman spectra of 242 lung adenocarcinoma cell samples and 231 normal cell samples are measured, followed by extracting the peak position and the intensity features. Furthermore, a multivariate correlation analysis of spectra features is performed to prevent the cross-sensitivity errors and to obtain a higher accuracy. Then, support vector machines with two different kernels, i.e., a polynomial function and a Gaussian radial basis function, are explored for the classification of all the samples. Benefiting from both the Raman spectra difference and the support vector machine, an accuracy of 97.5%, a sensitivity of 100%, and a specificity of 95% are achieved. The results indicate that this method is effective in identifying cancerous cells. Moreover, further work also demonstrates that the accuracy can reach 85%, even in the case of a small number of training samples (i.e., less than 100), so this method has great potential in the noninvasive detection of lung adenocarcinoma and can be easily applied to other fields of biophotonics in the future.