Markov-based photon statistics for blood flow estimation in parallelized diffuse correlation spectroscopy

We present what we believe to be a new computational framework for post-processing dynamic optical measurements, such as those arising in diffuse correlation spectroscopy (DCS), inspired by a Markov-based stochastic model and its associated statistics. Following detection of fluctuating low-light signals, DCS, along with related techniques, has historically computed a temporal autocorrelation to extract, e.g., particle motion. While well-known and widely used, the temporal autocorrelation exhibits several limitations. To address these, we develop an alternative transition probability approach to examine temporal statistics and experimentally apply it to parallelized DCS data from a 500x500 SPAD array. For high-count rate data, we find that around 90% of our signal of interest can be captured with 10× less the calculations that are typically performed via the temporal autocorrelation. These calculations are performed using photon count data selected from a time lag corresponding to the middle of the decorrelation curve computed in conventional DCS measurements. This improvement and pattern are observed in forearm blood flow measurements across a range of subjects.