Early-stage diagnosis of bladder cancer using surface-enhanced Raman spectroscopy combined with machine learning algorithms in a rat model.

Early diagnosis and accurate assessment of tumor development facilitate early bladder cancer resection and initiation of drug therapy. This study enabled an early, accurate, label-free, noninvasive diagnosis of bladder tumors by analyzing nano-biomarkers in a single drop of urine through surface-enhanced Raman spectroscopy (SERS). In a standard N-butyl-N-4-hydroxybutyl nitrosamine-induced rat model of bladder cancer, cancer stage and polyp tumor development were monitored using a small endoscope with a diameter of 1.2 mm in a minimally invasive manner without the need to kill the rats. Samples were divided into cancer-free, early-stage, and polyp-form cancer. Training data were classified according to micro-cystoscopic 5-aminolevulinic acid fluorescence diagnosis, and specimens were postmortem verified through histopathological analysis. A drop of urine from each sample group was placed on an Au-coated zinc oxide nanoporous chip to filter nano-biomaterials and selectively enhance the Raman signals of nanoscale analytes via SERS. Principal component analysis was used to reduce the dimensionality of the collected Raman spectra, and partial least squares discriminant analysis was used to find diagnostic clusters based on the labeled samples. The combination of SERS and machine learning achieved an accuracy ≥99.6% in diagnosing both early- and polyp-stage bladder tumors. With an area under the receiver operating characteristic curve greater than 0.996, the accuracy of the diagnosis in the rat model suggests that SERS-based diagnostic methods are promising when coupled with machine learning. Low-cost, label-free, and noninvasive surface-enhanced Raman spectra are ideal for developing clinically relevant point-of-care diagnostic techniques.