Detection of cystic and alveolar echinococcosis based on tissue surface-enhanced Raman spectroscopy combined with deep learning

Abstract

Echinococcosis chiefly includes cystic and alveolar echinococcosis, which is a parasitic disease. It is very important to find a quick and non-staining method to determine whether a tissue sample has echinococcosis lesions; it is not only conducive to the diagnosis of echinococcosis but also conducive to the judgment after surgery. In the study, tissue surface-enhanced Raman spectroscopy (SERS) in combination with deep learning was used to classify cystic and alveolar echinococcosis and healthy controls. Silver nanoparticles served as SERS-enhanced substrates, and a large amount of tissue SERS spectra was collected. There were 24 cases of cystic echinococcosis tissue, 14 cases of alveolar echinococcosis tissue, and 21 cases of healthy control tissues, and the numbers of SERS spectra collected were 594, 410, and 990, respectively, for a total of 1994 spectra. The convolutional neural network (CNN) was used to categorize SERS spectra into three types. Four other common machine learning classification algorithms were compared with the CNN model to highlight the classification effect of the CNN model. The results show that the model with the best effect is the CNN model, whose accuracy reaches 95%. Therefore, SERS combined with the CNN model has great potential for distinguishing the tissues of cystic and alveolar echinococcosis.