Spectral Barcode for Drug Classification Enabled with a Smartphone Raman Spectrometer

Abstract

Abstract Measuring, recoding and analyzing of spectral information of materials as its unique finger print using a ubiquitous smart phone has been craved for by scientists and consumers. We demonstrated the concept of a spectral barcode in drug classification with a smartphone Raman spectrometer. The spectral barcode is a 2D Raman spectral intensity map captured with a CMOS-image sensor integrated with a periodic array of band-pass filters. We classified the type of each drug by capturing its spectrum and using a convolutional neural network (CNN) that we trained to identify 11 major spectral components. This yielded a prediction accuracy of 99%, including drugs in the same category but from different brands. The accuracy of identifying drugs by individual brand was 79.5%, which improved to 83.2% simultaneously using the spectral barcode and RGB image in the CNN. This inherent property-based labeling system will facilitate fundamental research and business opportunities.