Cell classification with low-resolution Raman spectroscopy (LRRS)

The identification of individual eukaryotic and prokaryotic cells is the backbone of clinical pathology and provides crucial information about the genesis and progression of a disease. While most commonly fluorescent-label based methods are applied, label-free methods, such as Raman spectroscopy, are elegant alternatives. A major disadvantage of Raman spectroscopy is the low signal yield resulting in long acquisition times, making it impractical for high-throughput clinical analysis. As a rule, Raman-based cell identification relies on high-resolution Raman spectra. This comes at a cost of detected Raman photons. In this letter we show that while the proper biochemical characterization of cells requires high-resolution Raman spectra, the proper classification of cells does not. By varying the slit-width between 50 mu m and 500 mu m it is possible to show that detected Raman signal from eukaryotic cells increased up to seven-fold. Raman-based cell classification was performed on three cancer cell lines: Jurkat, MiaPaca2, and Capan1, at three different resolutions 8 cm(-1), 24 cm(-1), and 48 cm(-1). Moreover, we have simulated the resolution decrease due to low-diffraction gratings by binning neighboring pixels together. In both cases the cells were well classifiable using support vectors machine (SVM).