University of Edinburgh improves diffuse imaging of blood flow

10 May 2023

New data processing approach could relieve bottleneck for speckle techniques in clinics.

Diffuse correlation spectroscopy (DCS) can assess blood flow non-invasively, by analyzing diffused light returning from illuminated areas of tissue and detecting the speckled spectral signals of blood cells in motion.

The potential impact of DCS was recognized in a 2022 SPIE report, which concluded that "an exciting era of technology transfer is emerging as research groups have spun-out well-established, early-stage startup ventures intending to commercialize DCS for clinical use."

The SPIE report identified the increasing availability of advanced single-photon avalanche diode (SPAD) detectors as a key factor in the current rise of DCS techniques. However, those same detectors have introduced a potential new hurdle, caused by the increased data handling requirements of diffuse spectroscopic methods.

The extremely high data rates of modern SPAD cameras can exceed the maximum data transfer rates of commonly used communication protocols, a bottleneck that has limited the scalability of SPAD cameras to higher pixel resolutions and hindered the development of better multispeckle DCS techniques.

A project based at the University of Edinburgh and funded by Meta Platforms has now demonstrated a new data compression scheme that could improve the sensitivity and usability of multispeckle DCS instruments.

The study, published in Journal of Biomedical Optics, describes a novel data compression scheme in which most calculations involving SPAD data are performed directly on a commercial programmable circuit called a field-programmable gate array (FPGA). This alleviates the previous need for high computational power and extremely fast data transfer rates between the DSC system and the host system upon which the data is visualized, according to the project.

Clearer views of the brain

If the key part of the computational analysis, a per-pixel calculation termed the autocorrelation function, takes place locally on the FPGA, then a higher imaging frame rate can be maintained than is possible with existing hardware autocorrelators.

To test this approach, the Edinburgh project constructed a large array SPAD camera in which 128 linear autocorrelators were embedded in an FPGA integrated circuit. Packaged into a camera module christened Quanticam, this was able to calculate 12,288 channels of data and compute the ensemble autocorrelation function from 192 x 64 pixels of DCS data in real time.

"Our proposed system achieved a significant gain in the signal-to-noise ratio, which is 110 times higher than that possible on a single-speckle DSC implementation and 3 times higher than other state-of-the-art multispeckle DSC systems," commented Robert Henderson from the University of Edinburgh.

If FPGA-based designs can help researchers adopt SPAD arrays with high pixel resolution but without the data processing load currently involved, then SPAD cameras could become more widely adopted in the biomedical research community. This would expand the horizons of multispeckle DCS to more areas of biomedical research, including the imaging of cerebral blood dynamics.

"Intense research effort in SPAD camera development is currently ongoing to improve camera capabilities toward even larger pixel count, shorter exposure time and higher detection probability," said the project in its paper. "Soon we should expect high-performance SPAD cameras with FPGA-embedded or even on-chip computing that could surpass the multispeckle DCS requirements for noninvasive detection of local brain activation."