Microorganism detection using raman spectroscopy and C-ICA

Abstract

Microbial keratitis is an infection of the cornea that is caused by a variety of non-viral pathogens. It is the most potential complication of contact lens wear. Left untreated in time, it can cause serious damage to the eyes, to the point of rendering the patient blind. In this study, five sets of Raman Spectroscopy data were provided and processed using machine learning techniques. Principal Components-Linear Discriminant Analysis (PC- LDA) was first performed to classify the data and to obtain the accuracy of classifying each set of data. Next, Constrained Independent Component Analysis (C-ICA) was performed on the same datasets, and the correlation coefficient of the extracted signal was compared against the original signal. PC-LDA has been tested to be a proven technique in classifying the Raman spectra of the respective pure microorganism samples and the mixed microorganism samples on contact lens, but classification does not necessarily mean detection as there may be unknown contaminants in the sample. Synthetic data of the microorganism, namely P. Aeruginosa and C. Albicans, were successfully extracted from a source signal and it was shown that the C-ICA was able to detect the microorganism of interest on the surface of contact lens, even in low dosages. C-ICA has shown to be potential method of determining the presence of such pathogens, and the importance of which could lead to timely and appropriate treatment of microbial keratitis.