Deep learning of surface-enhanced Raman spectroscopy data for multiple sclerosis diagnostics

Despite the prevalence of multiple sclerosis (MS), there is currently no reliable biomarker for its identification. The existing diagnostic techniques either have high costs or lack specificity. Therefore, it is crucial to develop a diagnostic method that has high specificity and sensitivity and does not require complex sample processing or expensive equipment. The article demonstrates the application of a convolutional neural network (CNN) to the analysis of surface-enhanced Raman spectra of blood serum to distinguish between individuals with multiple sclerosis (MS) and healthy individuals. Additionally, it allows for the differentiation of patients based on the severity of their condition, as assessed through the use of the Expanded Disability Status Scale (EDSS). Through the implementation of CNN, we have achieved the ability to accurately differentiate between individuals with MS and healthy individuals with a specificity of 0.9, sensitivity, and accuracy of 1.0. Furthermore, the utilization of blood serum Raman spectra, combined with CNN, enables the categorization of patients according to their EDSS scores. The classification accuracy of the two groups (EDSS > 3.5 and EDSS <= 3.5) averaged 0.77. Overall, the study on the spectral properties of blood serum using surface-enhanced Raman spectroscopy represents a promising approach for diagnosing multiple sclerosis. Nevertheless, further in-depth investigations in this area are warranted.