Parallel high-throughput system for time domain diffuse optical spectroscopy based on a 16-channel SiPM array

The multi-channel SiPM technology is a fascinating leverage for time-resolved diffuse optical spectroscopy thanks to its remarkable parallelization capability that leads to rapidly measuring absorption and scattering properties of a turbid medium at multiple positions across a wide spectral range (600-1000 nm) at high throughput. For clinical applications, where the goal is to characterize the composition of biological tissues (e.g., fat, muscle, bone) in vivo non-invasively, these requirements are critical to support diagnosis with quantitative data, potentially reducing invasive procedures like biopsies and shortening waiting times for clinical exams. Therefore, we developed a time domain diffuse optical spectroscopy (TD-DOS) system based on a compact 16channel Silicon PhotoMultiplier (SiPM) array (footprint of 32 x 45 mm2, with single-photon timing resolution of 65 ps), capable of spectral or spatial parallelization. Spectral parallelization enables swift acquisition of extensive spectra, for example during functional tasks, allowing monitoring of task-related tissue changes and minimizing exam duration without compromising the informative content. Spatial parallelization facilitates tissue mapping or deep-layer investigation by leveraging the relationship between source-detector separation and penetration depth that holds when operating in the time domain. In this work, our system was configured to parallelize wavelengths across the 700-950 nm range (spectral resolution Delta lambda = 16 nm) to match key absorption peaks of hemoglobin and lipids, and the rising edge of water (peaking at 975 nm). Performance was evaluated applying the MEDPHOT protocol on phantoms, showing excellent linearity (worst R-2 = 0. 9973 for absorption and R-2 = 0. 9833 for reduced scattering), minimal absorption-scattering coupling, and remarkable absorption accuracy (average error of 3 % on absolute values), though scattering was overestimated (average error of + 17 %). In vivo trials demonstrated excellent reproducibility (CV < 5 % for absorption and < 4.5 % for scattering over 20 repetitions) and effective characterization of tissue during scans of the back and calf, correlating well with complementary ultrasound information about fat, muscle and bone layering. This system combines for the first time to our knowledge time domain insights, SiPM robustness, and parallelization speed, paving the way for efficient sample characterization in clinical and non-clinical contexts.