The detection of volatile organic compounds (VOCs) in urine is a noninvasive diagnostic technique with tremendous promise, holding great potential for early screening, auxiliary diagnosis, and progression monitoring of urinary system diseases. We developed a machine learning-assisted surface-enhanced Raman spectroscopy (SERS) strategy for the highly efficient enrichment and detection of urinary VOCs, aiming to achieve early screenings of chronic kidney disease (CKD). A composite aerogel, SMVF@GO@Ag, was fabricated as a SERS substrate, exhibiting a detection limit of 9.55 × 10-11 M for 4-MBN and a BET specific surface area of 5.57 m2/g. This substrate was integrated into a centrifuge tube to construct a VOC sensor, which was used for the enrichment of urinary VOCs and the collection of SERS spectral data. Subsequent application of machine learning for feature extraction and pattern recognition of the acquired SERS data set enabled, for the first time, the label-free collection of full-spectrum SERS data from urinary VOCs and the successful construction of a predictive model for CKD (with AUC > 0.95 on the ROC curve of DT, KNN, RF and SVM). Our study provides an innovative, highly sensitive, and low-cost solution with significant clinical translational value for advancing large-scale noninvasive screening of CKD.
Wang et al. (Thu,) studied this question.