Characterization of a Pixel-level Binarized Autocorrelator for Large Array Multispeckle Diffuse Correlation Spectroscopy under Photon-starved Conditions

Multispeckle Diffuse Correlation Spectroscopy (mDCS) is an advanced optical technology used to measure microvascular blood flow in deep tissue. It has emerged as a promising tool for continuous, real-time monitoring in clinical studies. However, its adoption in wearable applications is limited by the high resource demand for autocorrelation computation. To address this, we propose a resource-efficient 1-bit autocorrelator for in-pixel computation that exploits the inherent sparsity of photon detection events in long source-detector separation scenarios. By binarizing the detected photon counts, this architecture eliminates the need for multi-bit multipliers and reduces the bitwidth of shift registers. The proposed design is implemented on an FPGA and validated using real Single Photon Avalanche Diode (SPAD) inputs against a conventional 5-bit baseline under identical experimental conditions. It achieves 79% and 29.5% reductions in Look-Up Table and Flip-Flop usage, respectively. Systematic characterization using a rotating diffuser experimentally identifies an empirical threshold of 0.7 for the photon hit probability per lag bin, below which the 1-bit system estimates the decay time constant with tolerable error, closely matching the 5-bit baseline. The practical viability of this architecture is also demonstrated through a qualitative cuff occlusion measurement. This work presents a proof-of-concept for the 1-bit autocorrelator as a highly scalable solution toward massively parallel, large-array mDCS systems-on-chip for wearable healthcare devices.