Multiwavelength\nSurface-Enhanced Raman\nScattering Fingerprints of Human Urine for Cancer Diagnosis

Label-free surface-enhanced Raman spectroscopy (SERS)\nis capable\nof capturing rich compositional information from complex biosamples\nby providing vibrational spectra that are crucial for biosample identification.\nHowever, increasing complexity and subtle variations in biological\nmedia can diminish the discrimination accuracy of traditional SERS\nexcited by a single laser wavelength. Herein, we introduce a multiwavelength\nSERS approach combined with machine learning (ML)-based classification\nto improve the discrimination accuracy of human urine specimens for\nbladder cancer (BCa) diagnosis. This strategy leverages the excitation-wavelength-dependent\nSERS spectral profiles of complex matrices, which are mainly attributed\nto wavelength-related vibrational changes in individual analytes and\ndifferences in the variation ratios of SERS intensity across different\nwavelengths among various analytes. By capturing SERS fingerprints\nunder multiple excitation wavelengths, we can acquire more comprehensive\nand unique chemical information on complex samples. Further experimental\nexaminations with clinical urine specimens, supported by ML algorithms,\ndemonstrate the effectiveness of this multiwavelength strategy and\nimprove the diagnostic accuracy of BCa and staging of its invasion\nwith SERS spectra from increasing numbers of wavelengths. The multiwavelength\nSERS holds promise as a convenient, cost-effective, and broadly applicable\ntechnique for the precise identification of complex matrices and diagnosis\nof diseases based on body fluids.