Decoding motor plans using a closed-loop ultrasonic brain-machine interface.

Brain-machine interfaces (BMIs) enable people living with chronic paralysis to control computers, robots and more with nothing but thought. Existing BMIs have trade-offs across invasiveness, performance, spatial coverage and spatiotemporal resolution. Functional ultrasound (fUS) neuroimaging is an emerging technology that balances these attributes and may complement existing BMI recording technologies. In this study, we use fUS to demonstrate a successful implementation of a closed-loop ultrasonic BMI. We streamed fUS data from the posterior parietal cortex of two rhesus macaque monkeys while they performed eye and hand movements. After training, the monkeys controlled up to eight movement directions using the BMI. We also developed a method for pretraining the BMI using data from previous sessions. This enabled immediate control on subsequent days, even those that occurred months apart, without requiring extensive recalibration. These findings establish the feasibility of ultrasonic BMIs, paving the way for a new class of less-invasive (epidural) interfaces that generalize across extended time periods and promise to restore function to people with neurological impairments.

PerceptFlow: Real-Time Ultrafast Doppler Image Enhancement Using Deep Convolutional Neural Network and Perceptual Loss.

Ultrafast ultrasound is an emerging imaging modality derived from standard medical ultrasound. It allows for a high spatial resolution of 100 µm and a temporal resolution in the millisecond range with techniques such as ultrafast Doppler imaging. Ultrafast Doppler imaging has become a priceless tool for neuroscience, especially for visualizing functional vascular structures and navigating the brain in real time. Yet, the quality of a Doppler image strongly depends on experimental conditions and is easily subject to artifacts and deterioration, especially with transcranial imaging, which often comes at the cost of higher noise and lower sensitivity to small blood vessels. A common solution to better visualize brain vasculature is either accumulating more information, integrating the image over several seconds or using standard filter-based enhancement techniques, which often over-smooth the image, thus failing both to preserve sharp details and to improve our perception of the vasculature. In this study we propose combining the standard Doppler accumulation process with a real-time enhancement strategy, based on deep-learning techniques, using perceptual loss (PerceptFlow). With our perceptual approach, we bypass the need for long integration times to enhance Doppler images. We applied and evaluated our proposed method on transcranial Doppler images of mouse brains, outperforming state-of-the-art filters. We found that, in comparison to standard filters such as the Gaussian filter (GF) and block-matching and 3-D filtering (BM3D), PerceptFlow was capable of reducing background noise with a significant increase in contrast and contrast-to-noise ratio, as well as better preserving details without compromising spatial resolution.

Whole-Brain 3D Activation and Functional Connectivity Mapping in Mice using Transcranial Functional Ultrasound Imaging.

Functional ultrasound (fUS) imaging is a novel brain imaging modality that relies on the high-sensitivity measure of the cerebral blood volume achieved by ultrafast doppler angiography. As brain perfusion is strongly linked to local neuronal activity, this technique allows the whole-brain 3D mapping of task-induced regional activation as well as resting-state functional connectivity, non-invasively, with unmatched spatio-temporal resolution and operational simplicity. In comparison with fMRI (functional magnetic resonance imaging), a main advantage of fUS imaging consists in enabling a complete compatibility with awake and behaving animal experiments. Moreover, fMRI brain mapping in mice, the most used preclinical model in Neuroscience, remains technically challenging due to the small size of the brain and the difficulty to maintain stable physiological conditions. Here we present a simple, reliable and robust protocol for whole-brain fUS imaging in anesthetized and awake mice using an off-the-shelf commercial fUS system with a motorized linear transducer, yielding significant cortical activation following sensory stimulation as well as reproducible 3D functional connectivity pattern for network identification.

Mesoscopic and microscopic imaging of sensory responses in the same animal.

Imaging based on blood flow dynamics is widely used to study sensory processing. Here we investigated the extent to which local neuronal and capillary responses (two-photon microscopy) are correlated to mesoscopic responses detected with fast ultrasound (fUS) and BOLD-fMRI. Using a specialized chronic olfactory bulb preparation, we report that sequential imaging of the same mouse allows quantitative comparison of odour responses, imaged at both microscopic and mesoscopic scales. Under these conditions, functional hyperaemia occurred at the threshold of neuronal activation and fUS-CBV signals could be detected at the level of single voxels with activation maps varying according to blood velocity. Both neuronal and vascular responses increase non-linearly as a function of odour concentration, whereas both microscopic and mesoscopic vascular responses are linearly correlated to local neuronal calcium. These data establish strengths and limits of mesoscopic imaging techniques to report neural activity.

Vascular Compartmentalization of Functional Hyperemia from the Synapse to the Pia.

Functional hyperemia, a regional increase of blood flow triggered by local neural activation, is used to map brain activity in health and disease. However, the spatial-temporal dynamics of functional hyperemia remain unclear. Two-photon imaging of the entire vascular arbor in NG2-creERT2;GCaMP6f mice shows that local synaptic activation, measured via oligodendrocyte precursor cell (OPC) Ca2+ signaling, generates a synchronous Ca2+ drop in pericytes and smooth muscle cells (SMCs) enwrapping all upstream vessels feeding the activated synapses. Surprisingly, the onset timing, direction, and amplitude of vessel diameter and blood velocity changes vary dramatically from juxta-synaptic capillaries back to the pial arteriole. These results establish a precise spatial-temporal sequence of vascular changes triggered by neural activity and essential for the interpretation of blood-flow-based imaging techniques such as BOLD-fMRI.

Robust sound speed estimation for ultrasound-based hepatic steatosis assessment.

Hepatic steatosis is a common condition, the prevalence of which is increasing along with non-alcoholic fatty liver disease (NAFLD). Currently, the most accurate noninvasive imaging method for diagnosing and quantifying hepatic steatosis is MRI, which estimates the proton-density fat fraction (PDFF) as a measure of fractional fat content. However, MRI suffers several limitations including cost, contra-indications and poor availability. Although conventional ultrasound is widely used by radiologists for hepatic steatosis assessment, it remains qualitative and operator dependent. Interestingly, the speed of sound within soft tissues is known to vary slightly from muscle (1.575 mm · µs-1) to fat (1.450 mm · µs-1). Building upon this fact, steatosis could affect liver sound speed when the fat content increases. The main objectives of this study are to propose a robust method for sound speed estimation (SSE) locally in the liver and to assess its accuracy for steatosis detection and staging. This technique was first validated on two phantoms and SSE was assessed with a precision of 0.006 and 0.003 mm · µs-1 respectively for the two phantoms. Then a preliminary clinical trial (N = 17 patients) was performed. SSE results was found to be highly correlated with MRI proton density fat fraction (R 2 = 0.69) and biopsy (AUROC = 0.952) results. This new method based on the assessment of spatio-temporal properties of the local speckle noise for SSE provides an efficient way to diagnose and stage hepatic steatosis.

Light controls cerebral blood flow in naive animals.

Optogenetics is increasingly used to map brain activation using techniques that rely on functional hyperaemia, such as opto-fMRI. Here we test whether light stimulation protocols similar to those commonly used in opto-fMRI or to study neurovascular coupling modulate blood flow in mice that do not express light sensitive proteins. Combining two-photon laser scanning microscopy and ultrafast functional ultrasound imaging, we report that in the naive mouse brain, light per se causes a calcium decrease in arteriolar smooth muscle cells, leading to pronounced vasodilation, without excitation of neurons and astrocytes. This photodilation is reversible, reproducible and energy-dependent, appearing at about 0.5 mJ. These results impose careful consideration on the use of photo-activation in studies involving blood flow regulation, as well as in studies requiring prolonged and repetitive stimulations to correct cellular defects in pathological models. They also suggest that light could be used to locally increase blood flow in a controlled fashion.

Transcranial functional ultrasound imaging of the brain using microbubble-enhanced ultrasensitive Doppler.

Functional ultrasound (fUS) is a novel neuroimaging technique, based on high-sensitivity ultrafast Doppler imaging of cerebral blood volume, capable of measuring brain activation and connectivity in rodents with high spatiotemporal resolution (100µm, 1ms). However, the skull attenuates acoustic waves, so fUS in rats currently requires craniotomy or a thinned-skull window. Here we propose a non-invasive approach by enhancing the fUS signal with a contrast agent, inert gas microbubbles. Plane-wave illumination of the brain at high frame rate (500Hz compounded sequence with three tilted plane waves, PRF=1500Hz with a 128 element 15MHz linear transducer), yields highly-resolved neurovascular maps. We compared fUS imaging performance through the intact skull bone (transcranial fUS) versus a thinned-skull window in the same animal. First, we show that the vascular network of the adult rat brain can be imaged transcranially only after a bolus intravenous injection of microbubbles, which leads to a 9dB gain in the contrast-to-tissue ratio. Next, we demonstrate that functional increase in the blood volume of the primary sensory cortex after targeted electrical-evoked stimulations of the sciatic nerve is observable transcranially in presence of contrast agents, with high reproducibility (Pearson's coefficient ρ=0.7±0.1, p=0.85). Our work demonstrates that the combination of ultrafast Doppler imaging and injection of contrast agent allows non-invasive functional brain imaging through the intact skull bone in rats. These results should ease non-invasive longitudinal studies in rodents and open a promising perspective for the adoption of highly resolved fUS approaches for the adult human brain.

Multiplane wave imaging increases signal-to-noise ratio in ultrafast ultrasound imaging.

Ultrafast imaging using plane or diverging waves has recently enabled new ultrasound imaging modes with improved sensitivity and very high frame rates. Some of these new imaging modalities include shear wave elastography, ultrafast Doppler, ultrafast contrast-enhanced imaging and functional ultrasound imaging. Even though ultrafast imaging already encounters clinical success, increasing even more its penetration depth and signal-to-noise ratio for dedicated applications would be valuable. Ultrafast imaging relies on the coherent compounding of backscattered echoes resulting from successive tilted plane waves emissions; this produces high-resolution ultrasound images with a trade-off between final frame rate, contrast and resolution. In this work, we introduce multiplane wave imaging, a new method that strongly improves ultrafast images signal-to-noise ratio by virtually increasing the emission signal amplitude without compromising the frame rate. This method relies on the successive transmissions of multiple plane waves with differently coded amplitudes and emission angles in a single transmit event. Data from each single plane wave of increased amplitude can then be obtained, by recombining the received data of successive events with the proper coefficients. The benefits of multiplane wave for B-mode, shear wave elastography and ultrafast Doppler imaging are experimentally demonstrated. Multiplane wave with 4 plane waves emissions yields a 5.8 ± 0.5 dB increase in signal-to-noise ratio and approximately 10 mm in penetration in a calibrated ultrasound phantom (0.7 d MHz(-1) cm(-1)). In shear wave elastography, the same multiplane wave configuration yields a 2.07 ± 0.05 fold reduction of the particle velocity standard deviation and a two-fold reduction of the shear wave velocity maps standard deviation. In functional ultrasound imaging, the mapping of cerebral blood volume results in a 3 to 6 dB increase of the contrast-to-noise ratio in deep structures of the rodent brain.

Discriminative imaging of maternal and fetal blood flow within the placenta using ultrafast ultrasound.

Being able to map accurately placental blood flow in clinics could have major implications in the diagnosis and follow-up of pregnancy complications such as intrauterine growth restriction (IUGR). Moreover, the impact of such an imaging modality for a better diagnosis of placental dysfunction would require to solve the unsolved problem of discriminating the strongly intricated maternal and fetal vascular networks. However, no current imaging modality allows both to achieve sufficient sensitivity and selectivity to tell these entangled flows apart. Although ultrasound imaging would be the clinical modality of choice for such a problem, conventional Doppler echography both lacks of sensibility to detect and map the placenta microvascularization and a concept to discriminate both entangled flows. In this work, we propose to use an ultrafast Doppler imaging approach both to map with an enhanced sensitivity the small vessels of the placenta (~100 µm) and to assess the variation of the Doppler frequency simultaneously in all pixels of the image within a cardiac cycle. This approach is evaluated in vivo in the placenta of pregnant rabbits: By studying the local flow pulsatility pixel per pixel, it becomes possible to separate maternal and fetal blood in 2D from their pulsatile behavior. Significance Statement: The in vivo ability to image and discriminate maternal and fetal blood flow within the placenta is an unsolved problem which could improve the diagnosis of pregnancy complications such as intrauterine growth restriction or preeclampsia. To date, no imaging modality has both sufficient sensitivity and selectivity to discriminate these intimately entangled flows. We demonstrate that Ultrafast Doppler ultrasound method with a frame rate 100x faster than conventional imaging solves this issue. It permits the mapping of small vessels of the placenta (~100 µm) in 2D with an enhanced sensitivity. By assessing pixel-per-pixel pulsatility within single cardiac cycles, it achieves maternal and fetal blood flow discrimination.

Spatiotemporal Clutter Filtering of Ultrafast Ultrasound Data Highly Increases Doppler and fUltrasound Sensitivity.

Ultrafast ultrasonic imaging is a rapidly developing field based on the unfocused transmission of plane or diverging ultrasound waves. This recent approach to ultrasound imaging leads to a large increase in raw ultrasound data available per acquisition. Bigger synchronous ultrasound imaging datasets can be exploited in order to strongly improve the discrimination between tissue and blood motion in the field of Doppler imaging. Here we propose a spatiotemporal singular value decomposition clutter rejection of ultrasonic data acquired at ultrafast frame rate. The singular value decomposition (SVD) takes benefits of the different features of tissue and blood motion in terms of spatiotemporal coherence and strongly outperforms conventional clutter rejection filters based on high pass temporal filtering. Whereas classical clutter filters operate on the temporal dimension only, SVD clutter filtering provides up to a four-dimensional approach (3D in space and 1D in time). We demonstrate the performance of SVD clutter filtering with a flow phantom study that showed an increased performance compared to other classical filters (better contrast to noise ratio with tissue motion between 1 and 10mm/s and axial blood flow as low as 2.6 mm/s). SVD clutter filtering revealed previously undetected blood flows such as microvascular networks or blood flows corrupted by significant tissue or probe motion artifacts. We report in vivo applications including small animal fUltrasound brain imaging (blood flow detection limit of 0.5 mm/s) and several clinical imaging cases, such as neonate brain imaging, liver or kidney Doppler imaging.

Out-of-plane Doppler imaging based on ultrafast plane wave imaging.

Retrieving the out-of-plane blood flow velocity vector from two-dimensional transverse acquisitions of large vessels could improve the quantification of flow rate and maximum speed. The in-plane vector flow component can be computed easily using the Doppler frequency shift. The main problem is estimating the angle between the probe imaging plane and the vessel axis to derive the out-of-plane component from in-plane measurements. In this article, we study the case in which the velocity vector can be decomposed on two directions: the out-of-plane direction and the in-plane depth direction. We explore the combination of a technique called intrinsic spectral broadening with ultrafast plane wave imaging to retrieve the out-of-plane component of the flow velocity vector. Using a one-time probe calibration of this intrinsic spectral broadening, out-of-plane angle and flow speed can be recovered easily, thus avoiding approximations of a complex theoretical analysis. For the calibration step, ultrafast plane wave imaging permits a fast calibration procedure for the Doppler intrinsic spectral broadening. In vitro experimental validations are performed on a homogeneous flow phantom and a Poiseuille flow; the absolute speed was retrieved with 6% error. The potential of the technique is demonstrated in vivo on the human carotid artery. Combined with in-plane vector flow approaches, this out-of-plane Doppler imaging method paves the way to threedimensional vector flow imaging using only conventional onedimensional probe technology.

Functional ultrasound imaging of intrinsic connectivity in the living rat brain with high spatiotemporal resolution.

Long-range coherences in spontaneous brain activity reflect functional connectivity. Here we propose a novel, highly resolved connectivity mapping approach, using ultrafast functional ultrasound (fUS), which enables imaging of cerebral microvascular haemodynamics deep in the anaesthetized rodent brain, through a large thinned-skull cranial window, with pixel dimensions of 100 µm × 100 µm in-plane. The millisecond-range temporal resolution allows unambiguous cancellation of low-frequency cardio-respiratory noise. Both seed-based and singular value decomposition analysis of spatial coherences in the low-frequency (<0.1 Hz) spontaneous fUS signal fluctuations reproducibly report, at different coronal planes, overlapping high-contrast, intrinsic functional connectivity patterns. These patterns are similar to major functional networks described in humans by resting-state fMRI, such as the lateral task-dependent network putatively anticorrelated with the midline default-mode network. These results introduce fUS as a powerful novel neuroimaging method, which could be extended to portable systems for three-dimensional functional connectivity imaging in awake and freely moving rodents.

Transthoracic ultrafast Doppler imaging of human left ventricular hemodynamic function.

Heart diseases can affect intraventricular blood flow patterns. Real-time imaging of blood flow patterns is challenging because it requires both a high frame rate and a large field of view. To date, standard Doppler techniques can only perform blood flow estimation with high temporal resolution within small regions of interest. In this work, we used ultrafast imaging to map in 2-D human left ventricular blood flow patterns during the whole cardiac cycle. Cylindrical waves were transmitted at 4800 Hz with a transthoracic phased-array probe to achieve ultrafast Doppler imaging of the left ventricle. The high spatio-temporal sampling of ultrafast imaging permits reliance on a much more effective wall filtering and increased sensitivity when mapping blood flow patterns during the pre-ejection, ejection, early diastole, diastasis, and late diastole phases of the heart cycle. The superior sensitivity and temporal resolution of ultrafast Doppler imaging makes it a promising tool for the noninvasive study of intraventricular hemodynamic function.

Functional ultrasound imaging of the brain: theory and basic principles.

Hemodynamic changes in the brain are often used as surrogates of neuronal activity to infer the loci of brain activity. A major limitation of conventional Doppler ultrasound for the imaging of these changes is that it is not sensitive enough to detect the blood flow in small vessels where the major part of the hemodynamic response occurs. Here, we present a µDoppler ultrasound method able to detect and map the cerebral blood volume (CBV) over the entire brain with an important increase in sensitivity. This method is based on imaging the brain at an ultrafast frame rate (1 kHz) using compounded plane wave emissions. A theoretical model demonstrates that the gain in sensitivity of the µDoppler method is due to the combination of 1) the high signal-to-noise ratio of the gray scale images, resulting from the synthetic compounding of backscattered echoes; and 2) the extensive signal averaging enabled by the high temporal sampling of ultrafast frame rates. This µDoppler imaging is performed in vivo on trepanned rats without the use of contrast agents. The resulting images reveal detailed maps of the rat brain vascularization with an acquisition time as short as 320 ms per slice. This new method is the basis for a real-time functional ultrasound (fUS) imaging of the brain.

Aberration correction by time reversal of moving speckle noise.

Focusing a wave through heterogeneous media is an important problem in medical ultrasound imaging. In such aberrating media, in the presence of a small number of point reflectors, iterative time reversal is a well-known method able to focus on the strongest reflector. However, in presence of speckle noise generated by many non-resolved scatterers, iterative time reversal alone does not work. In this paper, we propose the use of the echoes coming from moving particles in a flow, such as red blood cells, to generate a virtual point reflector by iterative time reversal. The construction of the virtual point reflector is performed by a coherent addition of independent realizations of speckle coming from moving particles. After focusing on a virtual point reflector, ultrasound images can be locally corrected inside an isoplanatic patch. An application for the correction of power Doppler images is presented. A theoretical analysis shows that this iterative method allows focusing on the point of maximal insonification of the uncorrected beam.

Ultrafast Doppler imaging of blood flow dynamics in the myocardium.

Imaging intramyocardial vascular flows in real-time could strongly help to achieve better diagnostic of cardiovascular diseases. To date, no standard imaging modality allows describing accurately myocardial blood flow dynamics with good spatial and temporal resolution. We recently introduced a novel ultrasonic Doppler imaging technique based on compounded plane waves transmissions at ultrafast frame rate. The high sensitivity of this ultrafast Doppler technique permits to image the intramyocardial blood flow and its dynamics. A dedicated demodulation-filtering process is implemented to compensate for the large tissue velocity of the myocardium during the cardiac cycle. A signed power Doppler processing provides the discrimination between arterial and venous flows. Experiments were performed in vivo in a large animal open chest model ( N = 5 sheep) using a conventional ultrasonic probe placed at the surface of the heart. Results show the capability of the technique to image intramyocardial vascular flows in normal physiological conditions with good spatial (200 µm) and temporal resolution (10 ms). Flow dynamics over the cardiac cycle were investigated and the imaging method demonstrated a phase opposition of flow waveforms between arterial and venous flows. Finally, ultrafast Doppler combined with tissue motion compensation was found able to reveal vascular flow disruption in ischemic regions during occlusion of the main diagonal coronary artery.