Deep-learning-based separation of shallow and deep layer blood flow rates in diffuse correlation spectroscopy

Beneath the surface: revealing deep-tissue blood flow in human subjects with massively parallelized diffuse correlation spectroscopy

Significance

Diffuse correlation spectroscopy (DCS) allows label-free, non-invasive investigation of microvascular dynamics deep within tissue, such as cerebral blood flow (CBF). However, the signal-to-noise ratio (SNR) in DCS limits its effective cerebral sensitivity in adults, in which the depth to the brain, through the scalp and skull, is substantially larger than in infants.

Aim

Therefore, we aim to increase its SNR and, ultimately, its sensitivity to CBF through new DCS techniques.

Approach

We present an in vivo demonstration of parallelized DCS (PDCS) to measure cerebral and muscular blood flow in healthy adults. Our setup employs an innovative array with hundreds of thousands single photon avalanche diodes (SPAD) in a 500 × 500 grid to boost SNR by averaging all independent pixel measurements. We tested this device on different total pixel counts and frame rates. A secondary, smaller array was used for reference measurements from shallower tissue at lower source-detector-separation (SDS).

Results

The new system can measure pulsatile blood flow in cerebral and muscular tissue, at up to 4 cm SDS, while maintaining a similar measurement noise as compared with a previously published 32 × 32 PDCS system at 1.5 cm SDS. Data from a cohort of 15 adults provide strong experimental evidence for functional CBF activity during a cognitive memory task and allowed analysis of pulse markers. Additional control experiments on muscular blood flow in the forearm with a different technical configuration provide converging evidence for the efficacy of this technique.

Conclusions

Our results outline successful PDCS measurements with large SPAD arrays to enable detect CBF in human adults. The ongoing development of SPAD camera technology is expected to result in larger and faster detectors in the future. In combination with new data processing techniques, tailored for the sparse signal of binary photon detection events in SPADs, this could lead to even greater SNR increase and ultimately greater depth sensitivity of PDCS.

Fast blood flow index reconstruction of diffuse correlation spectroscopy using a back-propagation-free data-driven algorithm

This study introduces a fast and accurate online training method for blood flow index (BFI) and relative BFI (rBFI) reconstruction in diffuse correlation spectroscopy (DCS). We implement rigorous mathematical models to simulate the auto-correlation functions (g 2) for semi-infinite homogeneous and three-layer human brain models. We implemented a fast online training algorithm known as random vector functional link (RVFL) to reconstruct BFI from noisy g 2. We extensively evaluated RVFL regarding both speed and accuracy for training and inference. Moreover, we compared RVFL with extreme learning machine (ELM) architecture, a conventional convolutional neural network (CNN), and three fitting algorithms. Results from semi-infinite and three-layer models indicate that RVFL achieves higher accuracy than the other algorithms, as evidenced by comprehensive metrics. While RVFL offers comparable accuracy to CNNs, it boosts training speeds that are 3900-fold faster and inference speeds that are 19.8-fold faster, enhancing its generalizability across different experimental settings. We also used g 2 from one- and three-layer Monte Carlo (MC)-based in-silico simulations, as well as from analytical models, to compare the accuracy and consistency of the results obtained from RVFL and ELM. Furthermore, we discuss how RVFL is more suitable for embedded hardware due to its lower computational complexity than ELM and CNN for training and inference.

A comprehensive overview of diffuse correlation spectroscopy: Theoretical framework, recent advances in hardware, analysis, and applications

Diffuse correlation spectroscopy (DCS) is a powerful tool for assessing microvascular hemodynamic in deep tissues. Recent advances in sensors, lasers, and deep learning have further boosted the development of new DCS methods. However, newcomers might feel overwhelmed, not only by the already-complex DCS theoretical framework but also by the broad range of component options and system architectures. To facilitate new entry to this exciting field, we present a comprehensive review of DCS hardware architectures (continuous-wave, frequency-domain, and time-domain) and summarize corresponding theoretical models. Further, we discuss new applications of highly integrated silicon single-photon avalanche diode (SPAD) sensors in DCS, compare SPADs with existing sensors, and review other components (lasers, sensors, and correlators), as well as data analysis tools, including deep learning. Potential applications in medical diagnosis are discussed and an outlook for the future directions is provided, to offer effective guidance to embark on DCS research.

A Device-on-Chip Solution for Real-Time Diffuse Correlation Spectroscopy Using FPGA

Diffuse correlation spectroscopy (DCS) is a non-invasive technology for the evaluation of blood perfusion in deep tissue. However, it requires high computational resources for data analysis, which poses challenges in its implementation for real-time applications. To address the unmet need, we developed a novel device-on-chip solution that fully integrates all the necessary computational components needed for DCS. It takes the output of a photon detector and determines the blood flow index (BFI). It is implemented on a field-programmable gate array (FPGA) chip including a multi-tau correlator for the calculation of the temporal light intensity autocorrelation function and a DCS analyzer to perform the curve fitting operation that derives the BFI at a rate of 6000 BFIs/s. The FPGA DCS system was evaluated against a lab-standard DCS system for both phantom and cuff ischemia studies. The results indicate that the autocorrelation of the light correlation and BFI from both the FPGA DCS and the reference DCS matched well. Furthermore, the FPGA DCS system was able to achieve a measurement rate of 50 Hz and resolve pulsatile blood flow. This can significantly lower the cost and footprint of the computational components of DCS and pave the way for portable, real-time DCS systems.

Correlation of diabetic renal hypoperfusion with microvascular responses of the skeletal muscle: a rat model study using diffuse correlation spectroscopy

Using diffuse correlation spectroscopy, we assessed the renal blood flow and thigh muscle microvascular responses in a rat model of type 2 diabetes. The blood flow index at the renal surface decreased significantly with arterial clamping, cardiac extirpation, and the progression of diabetic endothelial dysfunction. Renal blood flow measured in diabetic and nondiabetic rats also showed a significant correlation with the reactive hyperemic response of the thigh muscle. These results suggest shared microcirculatory dysfunction in the kidney and skeletal muscle and support endothelial responses in the skeletal muscle as a potential noninvasive biomarker of renal hypoperfusion.

Differential changes in blood flow and oxygen utilization in active muscles between voluntary exercise and electrical muscle stimulation in young adults

The physiological effects on blood flow and oxygen utilization in active muscles during and after involuntary contraction triggered by electrical muscle stimulation (EMS) remain unclear, particularly compared with those elicited by voluntary (VOL) contractions. Therefore, we used diffuse correlation and near-infrared spectroscopy (DCS-NIRS) to compare changes in local muscle blood flow and oxygen consumption during and after these two types of muscle contractions in humans. Overall, 24 healthy young adults participated in the study, and data were successfully obtained from 17 of them. Intermittent (2-s contraction, 2-s relaxation) isometric ankle dorsiflexion with a target tension of 20% of maximal VOL contraction was performed by EMS or VOL for 2 min, followed by a 6-min recovery period. DCS-NIRS probes were placed on the tibialis anterior muscle, and relative changes in local tissue blood flow index (rBFI), oxygen extraction fraction (rOEF), and metabolic rate of oxygen (rMRO2) were continuously derived. EMS induced more significant increases in rOEF and rMRO2 than VOL exercise but a comparable increase in rBFI. After EMS, rBFI and rMRO2 decreased more slowly than after VOL and remained significantly higher until the end of the recovery period. We concluded that EMS augments oxygen consumption in contracting muscles by enhancing oxygen extraction while increasing oxygen delivery at a rate similar to the VOL exercise. Under the conditions examined in this study, EMS demonstrated a more pronounced and/or prolonged enhancement in local muscle perfusion and aerobic metabolism compared with VOL exercise in healthy participants.NEW & NOTEWORTHY This is the first study to visualize continuous changes in blood flow and oxygen utilization within contracted muscles during and after electrical muscle stimulation (EMS) using combined diffuse correlation and near-infrared spectroscopy. We found that initiating EMS increases blood flow at a rate comparable to that during voluntary (VOL) exercise but enhances oxygen extraction, resulting in higher oxygen consumption. Furthermore, EMS increased postexercise muscle perfusion and oxygen consumption compared with that after VOL exercise.