Abstract

Raman spectroscopy is a powerful analytical tool to be used in many biomedical applications and could be potentially translated into clinical work. The challenge of Raman spectroscopy in biomedical applications is the high inherent fluorescence of biological samples. One promising method to suppress the fluorescence background is to use pulsed lasers and time-gated detectors but the complexity of time-gated systems has hindered their widespread usage. We present here chemical imaging of human teeth by means of a new kind of compact and practical fluorescence-suppressed Raman spectrometer based on a time-resolved 16 × 256 CMOS single-photon avalanche diode (SPAD) line sensor with an integrated 256-channel 3-bit on-chip time-to-digital converter. The chemical images were constructed by utilizing a simple unsupervised machine learning algorithm (k-means clustering). The high quality of Raman spectra measured with the time-resolved CMOS SPAD-based Raman spectrometer was verified by comparing the spectra to those collected with a commercial conventional continuous wave (CW) Raman spectrometer. The spectra measured by using the time-resolved CMOS SPAD-based Raman spectrometer had 4.4–8.8 times higher signal to peak-to-peak noise ratio values than the spectra from the CW Raman spectrometer when the same radiant exposure (∼300 J mm−2) was used with both spectrometers. This paper shows in practice the potential of time-resolved CMOS SPAD-based Raman spectroscopy in the field of biomedicine and we expect that the presented technology could pave the way for the development of new kind of compact and practical fluorescence-suppressed Raman spectrometers to be used both in biomedical research and clinical settings.