A biosensing method for the direct serological detection of liver diseases by integrating a SERS-based sensor and a CNN classifier

Direct serological detection, due to its clinical facility and testing economy, affords prominent clinical values to the early detection of cancer. Surface-enhanced Raman spectroscopy (SERS)-based sensors have shown great promise in realizing this form of detection. Detecting liver cancer early with such a form, especially in terms of monitoring the pathogenic progression from hepatic inflammations to cancer, is the most effective clinical path to reducing the mortality rate. However, the methodology investigation for this purpose remains a formidable challenge. We fabricated a SERS-based sensor, consisting of Au-Ag nanocomplex-decorated ZnO nanopillars on paper. The sensor has an analytic enhancement factor of 1.02 x 107, which is enough to sense the biomolecular information of liver diseases through direct serum SERS analysis. A convolutional neural network (CNN) classifier for recognizing serum SERS spectra was constructed by deep learning. Integrating this sensor with the CNN, we established an intelligent biosensing method and realized direct serological detection of liver diseases within 1 min. As a proof-of-concept, the method achieved a prediction accuracy of 97.78% on an independent test dataset randomly sampled from 30 normal controls, 30 hepatocellular carcinoma (HCC) cases, and 30 hepatitis B (HB) patients. The results suggest this method can be developed for detecting liver diseases clinically and is worthy of exploration as a means of liver cancer surveillance. The presented sensor holds potential for clinical translation to the direct serological detection of diseases.