Deep learning model for ultrafast quantification of blood flow in diffuse correlation spectroscopy

Quantification of blood flow index in diffuse correlation spectroscopy using a robust deep learning method

Significance

Diffuse correlation spectroscopy (DCS) is a powerful, noninvasive optical technique for measuring blood flow. Traditionally the blood flow index (BFi) is derived through nonlinear least-square fitting the measured intensity autocorrelation function (ACF). However, the fitting process is computationally intensive, susceptible to measurement noise, and easily influenced by optical properties (absorption coefficient μ a and reduced scattering coefficient μ s ' ) and scalp and skull thicknesses.

Aim

We aim to develop a data-driven method that enables rapid and robust analysis of multiple-scattered light's temporal ACFs. Moreover, the proposed method can be applied to a range of source-detector distances instead of being limited to a specific source-detector distance.

Approach

We present a deep learning architecture with one-dimensional convolution neural networks, called DCS neural network (DCS-NET), for BFi and coherent factor (β ) estimation. This DCS-NET was performed using simulated DCS data based on a three-layer brain model. We quantified the impact from physiologically relevant optical property variations, layer thicknesses, realistic noise levels, and multiple source-detector distances (5, 10, 15, 20, 25, and 30 mm) on BFi and β estimations among DCS-NET, semi-infinite, and three-layer fitting models.

Results

DCS-NET shows a much faster analysis speed, around 17,000-fold and 32-fold faster than the traditional three-layer and semi-infinite models, respectively. It offers higher intrinsic sensitivity to deep tissues compared with fitting methods. DCS-NET shows excellent anti-noise features and is less sensitive to variations of μ a and μ s ' at a source-detector separation of 30 mm. Also, we have demonstrated that relative BFi (rBFi) can be extracted by DCS-NET with a much lower error of 8.35%. By contrast, the semi-infinite and three-layer fitting models result in significant errors in rBFi of 43.76% and 19.66%, respectively.

Conclusions

DCS-NET can robustly quantify blood flow measurements at considerable source-detector distances, corresponding to much deeper biological tissues. It has excellent potential for hardware implementation, promising continuous real-time blood flow measurements.

Comparison of numerical-integration-based methods for blood flow estimation in diffuse correlation spectroscopy

BACKGROUND AND OBJECTIVE

Diffuse correlation spectroscopy (DCS) is an optical blood flow monitoring technology that has been utilized in various biomedical applications. In signal processing of DCS, nonlinear fitting of the experimental data and the theoretical model can be a hindrance in real-time blood flow monitoring. As one of the approaches to resolve the issue, INISg1, the inverse of numerical integration of squared g1 (a normalized electric field autocorrelation function), that could surpass the state-of-the-art technique at the time in terms of signal processing speed, has been introduced. While it is possible to implement INISg1 using various numerical integration methods, no relevant studies have been performed. Meanwhile, INISg1 was only tested within limited experimental conditions, which cannot guarantee the robustness of INISg1 in various experimental conditions. Thus, this study aims to introduce variants of INISg1 and perform a thorough comparison of the original INISg1 and its variants.

METHODS

In this study, based on the right Riemann sum (RR) and trapezoid rule (TR) of numerical integration, INISg1_RR and INISg1_TR are suggested. They are thoroughly compared with the original INISg1 using model-based simulations that offer us control of most of the experimental conditions, including integration time, β, and photon count rate.

RESULTS

Except for some extreme cases, INISg1 performed more robustly than INISg1_RR and INISg1_TR. However, in extreme conditions, variants of INISg1 performed better than INISg1. With the same condition, the signal processing speed of INISg1 was 1.63 and 1.98 times faster than INISg1_RR and INISg1_TR, respectively.

CONCLUSION

This study shows that INISg1 is robust in most cases and the study can be a guide for researchers using INISg1 and its variants in different types of DCS applications.

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

Diffuse correlation spectroscopy faces challenges concerning the contamination of cutaneous and deep tissue blood flow. We propose a long short-term memory network to directly quantify the flow rates of shallow and deep-layer tissues. By exploiting the different contributions of shallow and deep-layer flow rates to auto-correlation functions, we accurately predict the shallow and deep-layer flow rates (RMSE = 0.047 and 0.034 ml/min/100 g of simulated tissue, R2 = 0.99 and 0.99, respectively) in a two-layer flow phantom experiment. This approach is useful in evaluating the blood flow responses of active muscles, where both cutaneous and deep-muscle blood flow increase with exercise.

Current capabilities and future perspectives of FCS: super-resolution microscopy, machine learning, and in vivo applications

Fluorescence correlation spectroscopy (FCS) is a single molecule sensitive tool for the quantitative measurement of biomolecular dynamics and interactions. Improvements in biology, computation, and detection technology enable real-time FCS experiments with multiplexed detection even in vivo. These new imaging modalities of FCS generate data at the rate of hundreds of MB/s requiring efficient data processing tools to extract information. Here, we briefly review FCS's capabilities and limitations before discussing recent directions that address these limitations with a focus on imaging modalities of FCS, their combinations with super-resolution microscopy, new evaluation strategies, especially machine learning, and applications in vivo.

Blood flow estimation via numerical integration of temporal autocorrelation function in diffuse correlation spectroscopy

BACKGROUND AND OBJECTIVE

Diffuse correlation spectroscopy (DCS) is an optical technique widely used to monitor blood flow. Recently, efforts have been made to derive new signal processing methods to minimize the systems used and shorten the signal processing time. Herein, we propose alternative approaches to obtain blood flow information via DCS by numerically integrating the temporal autocorrelation curves.

METHODS

We use the following methods: the inverse of K² (IK2)-based on the framework of diffuse speckle contrast analysis-and the inverse of the numerical integration of squared g₁ (INISg1) which, based on the normalized electric field autocorrelation curve, is more simplified than IK2. In addition, g₁ thresholding is introduced to further reduce computational time and make the suggested methods comparable to the conventional nonlinear fitting approach. To validate the feasibility of the suggested methods, studies using simulation, liquid phantom, and in vivo settings were performed. In the meantime, the suggested methods were implemented and tested on three types of Arduino (Arduino Due, Arduino Nano 33 BLE Sense, and Portenta H7) to demonstrate the possibility of miniaturizing the DCS systems using microcotrollers for signal processing.

RESULTS

The simulation and experimental results confirm that both IK2 and INISg1 are sufficiently relevant to capture the changes in blood flow information. More interestingly, when g₁ thresholding was applied, our results showed that INISg1 outperformed IK2. It was further confirmed that INISg1 with g₁ thresholding implemented on a PC and Portenta H7, an advanced Arduino board, performed faster than did the deep learning-based, state-of-the-art processing method.

CONCLUSION

Our findings strongly indicate that INISg1 with g₁ thresholding could be an alternative approach to derive relative blood flow information via DCS, which may contribute to the simplification of DCS methodologies.

Performance optimisation of a holographic Fourier domain diffuse correlation spectroscopy instrument

We have previously demonstrated a novel interferometric multispeckle Fourier domain diffuse correlation spectroscopy system that makes use of holographic camera-based detection, and which is capable of making in vivo pulsatile flow measurements. In this work, we report on a systematic characterisation of the signal-to-noise ratio performance of our system. This includes demonstration and elimination of laser mode hopping, and correction for the instrument's modulation transfer function to ensure faithful reconstruction of measured intensity profiles. We also demonstrate a singular value decomposition approach to ensure that spatiotemporally correlated experimental noise sources do not limit optimal signal-to-noise ratio performance. Finally, we present a novel multispeckle denoising algorithm that allows our instrument to achieve a signal-to-noise ratio gain that is equal to the square root of the number of detected speckles, whilst detecting up to ∼1290 speckles in parallel. The signal-to-noise ratio gain of 36 that we report is a significant step toward mitigating the trade-off that exists between signal-to-noise ratio and imaging depth in diffuse correlation spectroscopy.

Deep Learning in Biomedical Optics

This article reviews deep learning applications in biomedical optics with a particular emphasis on image formation. The review is organized by imaging domains within biomedical optics and includes microscopy, fluorescence lifetime imaging, in vivo microscopy, widefield endoscopy, optical coherence tomography, photoacoustic imaging, diffuse tomography, and functional optical brain imaging. For each of these domains, we summarize how deep learning has been applied and highlight methods by which deep learning can enable new capabilities for optics in medicine. Challenges and opportunities to improve translation and adoption of deep learning in biomedical optics are also summarized. Lasers Surg. Med. © 2021 Wiley Periodicals LLC.

Quantification of blood flow index in diffuse correlation spectroscopy using long short-term memory architecture

Diffuse correlation spectroscopy (DCS) is a noninvasive technique that derives blood flow information from measurements of the temporal intensity fluctuations of multiply scattered light. Blood flow index (BFI) and especially its variation was demonstrated to be approximately proportional to absolute blood flow. We investigated and assessed the utility of a long short-term memory (LSTM) architecture for quantification of BFI in DCS. Phantom and in vivo experiments were established to measure normalized intensity autocorrelation function data. Improved accuracy and faster computational time were gained by the proposed LSTM architecture. The results support the notion of using proposed LSTM architecture for quantification of BFI in DCS. This approach would be especially useful for continuous real-time monitoring of blood flow.