| abstract |
Disclosed herein are substrates for surface-enhanced Raman spectroscopy (SERS), methods of fabrication of the same using soft and nanoparticle lithography or laser-induced nano structuring of thin films (LINST), and their use to characterize extracellular vesicles (EVs) from a range of sources including but not limited to cancers, bacteria, viruses and/or placental cells. Also disclosed are machine learning methods for classifying and/or identifying SERS spectra from particles including EVs, the machine learning methods including bottleneck classifiers or layers configured to reduce the dimension of the network. In further methods the bottleneck classifier is combined with an autoencoder in either a supervised or unsupervised manner to identify of classify SERs spectra features. |