Rapid detection of Pseudomonas aeruginosa based on lab-on-a-chip platform using immunomagnetic separation, light scattering, and machine learning

Abstract

The rapid detection of the pathogenic bacteria in patient samples is crucial to expedient patient care. The proposed approach reports the development of a novel lab-on-a-chip device for the rapid detection of P. aeruginosa based on immunomagnetic separation, optical scattering, and machine learning. The immunomagnetic particles with a diameter of 5 μm were synthesized for isolating P. aeruginosa from the test sample. A microfluidic chip was fabricated, and three optical fibers were embedded for connecting a laser light and two photodetectors. The laser light was pointed towards the channel to pass light through the sample. A pair of photodetectors via optical fibers were arranged symmetrically at 45° to the channel. The photodetectors acquired scattered light from the flowing sample and converted the light to an electrical signal. The sample containing immunomagnetic beads linked with bacteria was injected into the microfluidic chip. The optimized conditions for performing the experiments were characterized for real-time detection of P. aeruginosa. The data acquisition system recorded the real-time light scattering from the test sample. After removing noise from the output waveform, five different time-domain statistical features were extracted from each waveform: standard mean, standard variance, skewness, kurtosis, and coefficient of variation. The pathogens classification was performed by training the discrimination model using extracted features based on machine learning algorithms. The support vector machines (SVM) with a sigmoid function kernel showed superior classification performance with 97.9% accuracy among other classifiers, including k-nearest neighbors (KNN), logistic regression (LR), and naïve Bayes (NB). The method can detect P. aeruginosa specifically and quantitatively with a limit of detection of 102 CFU/mL. The device can classify P. aeruginosa within 10 min with a total assay time of 25 min. The device was used to test its ability to detect pathogen from the serum and sputum specimens spiked with 105 CFU/mL concentration of P. aeruginosa. The results indicate that light scattering combined with machine learning can be used to detect P. aeruginosa. The proposed technique is anticipated to be helpful as a rapid device for diagnosing P. aeruginosa related infections.