Digital retina combined with SERS technology to capture hepatobiliary biomarkers in serum

Hepatobiliary diseases pose significant challenges for early clinical detection due to their subtle initial symptoms. This study combines Ag@5/Ag-Cu@20/PSS surface-enhanced Raman scattering (SERS) sensors with “digital retina” technology to establish an intelligent diagnostic strategy for automatically identifying serum biomarkers. By collecting peripheral serum samples from healthy individuals and patients with hepatobiliary diseases, SERS spectra were obtained using the Ag@5/Ag-Cu@20/PSS sensor. Feature peak analysis revealed significant differences between the healthy group and the patient group in multiple SERS feature peaks, reflecting disease-related biochemical changes. After preprocessing, various deep learning models were employed for classification evaluation. The results showed that the “digital retina” framework centered on a multi-layer perceptron (MLP) achieved an accuracy of 92.34 % on the test set, with an area under the ROC curve (AUC) value of 0.9738, outperforming models such as deep neural networks (DNN), ResNet, simple convolutional neural networks (SimpleCNN), AlexNet, and Transformers. Five-fold cross-validation further validated the model’s robustness and generalization ability. This study demonstrates that the combination of SERS platforms with deep learning enables highly sensitive, non-invasive, and automated detection of serum biomarkers for hepatobiliary diseases. The “digital retina” technology holds significant potential for early diagnosis, health screening, and precision medicine management.