Approach for Label-Free and High-Accuracy Identification of Alzheimer’s Disease Combining Machine Learning and SERS Technology

Alzheimer’s disease (AD) is a fatal neurodegenerative disorder that is difficult to identify early in clinical practice. To overcome the existing limitations, we designed a label-free and highly accurate serum analysis platform for identifying AD in rats. Specifically, the closely ordered Nano Stellar Array (NASA) substrate was prepared to measure the serum of Aβ1–40-induced AD rats. Next, a recognition model was constructed using the principal component analysis (PCA)–linear discriminant analysis (LDA) algorithm. The prepared NASA substrate exhibited uniformity, high stability, reproducibility, high sensitivity, and cleanliness, enabling the acquisition of high-quality surface-enhanced Raman spectroscopy (SERS) spectra. The PCA-LDA model demonstrated good performance in differentiating the SERS spectra of serum from AD rats, with an accuracy of 95.8%, a sensitivity of 100%, a specificity of 92%, and an area under curve (AUC) of 0.97. This study not only provides a practical platform for label-free serum analysis to identify AD but also presents an approach for serum profiling at the spectral level.