Improving the performance of the echinococcosis diagnosis model based on serum Raman spectroscopy via the integration of convolutional neural network and support vector machine.

Echinococcosis is a zoonotic parasitic disease characterized by its insidious nature and severe health impacts. Rapid and accurate screening is crucial for subsequent treatment. Previous studies have demonstrated that Raman spectroscopy combined with machine learning or deep learning can be used for rapid diagnosis of echinococcosis, but there remains room for improving diagnostic accuracy. Therefore, this study proposes combining a convolutional neural network with a support vector machine (CNN-SVM) for the analysis of serum Raman spectra, aiming to achieve high-accuracy classification of echinococcosis, liver cirrhosis, hepatocellular carcinoma, and normal control groups. After collecting the spectra of 573 serum samples, spectral features were extracted by the CNN and then classified using the SVM. The results show that the classification accuracy of CNN-SVM model is 96.5 %, which is better than the CNN (92.3 %) and SVM (89.3 %) used alone. Furthermore, in the binary classification task of detecting echinococcosis versus non-echinococcosis cases, the CNN-SVM model also achieved an accuracy of 96.5 %, surpassing the traditional dot immunogold filtration assay (88.7 %). In conclusion, the proposed CNN-SVM model demonstrates superior diagnostic performance for echinococcosis and holds significant clinical application potential.