functional UltraSound (fUS) imaging
An innovative technology for imaging brain activity with a high spatiotemporal resolution (80 µm/20 ms)
Fast, Cheap, Non radiative, No contrast agent, Imaging in depth, Compatible with chronic studies
Brain imaging includes different methods that fall into two broad categories: structural and functional imaging. Structural imaging investigates the structure of the brain and can be used for the diagnosis of large-scale intracranial diseases such as tumors, and injuries. Functional imaging reveals the activity in certain brain regions by detecting changes in metabolism, blood flow, regional chemical composition and/or absorption. The major challenge in the years ahead will be to combine both real-time and non-invasive brain imaging.
Brain imaging technologies are crucial for understanding the relationships between specific areas of the brain and their function, helping to locate the areas of the brain that are affected by diseases or neurological disorders and build new strategies to treat them.
Spatio-temporal resolution of the main functional imaging techniques (adapted from Monet@Yonsei)
Thanks to the innovative approach developed by our multidisciplinary team, this new brain imaging system is no longer a dream and our hope is that it will be available to everyone in the near future. Our breakthrough technology was thoroughly built up with the goal of offering superior products for neuroscientists and medical doctors. For the first time, it is not only possible to acquire high-resolution images of the brain vasculature in a few milliseconds but also to record hemodynamics of the entire brain, in real-time.
This technology offers researchers a set of unique features and benefits, such as a new micro-Doppler mode that we termed functional UltraSound (fUS). fUS has been designed to become a centerpiece in both preclinical and clinical studies as it can be used to diagnose metabolic diseases and lesions on a finer scale, for neurological and cognitive psychological research as well as brain-computer interfaces.
I. Functional UltraSound (fUS) Imaging
Ultrasound has been used for diagnostic purposes for more than 20 years in almost all branches of medicine and is widely accepted due to its clinical usefulness, convenience and non-invasiveness. The use of ultrasound as an effective preclinical modality also continues to increase worldwide. Ultrasound is an early technique developed in the 1930s and 40s that was primarily used neurologically until the 1960s to try to identify brain tumors. But, when scientists determined that the skull significantly distorts and attenuates the signals, its use for this purpose stopped.
Structural and functional imaging using ultrasound. Credits OBI-Team.
Fortuitously, abandonment of ultrasound to try to detect brain tumors came at the same time that new radiological technologies for brain imaging such as Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) were emerging. These powerful modalities relegate ultrasound brain imaging to specific applications such as monitoring of cerebral blood flow in patients with severe head trauma or neonatal imaging through the fontanel window. Until recently, ultrasound imaging was unable to produce functional images of the brain because Doppler ultrasound is not sensitive enough to detect small blood flow changes.
However, functional UltraSound allows recording of specific brain areas activated during a certain stimulus or function. This arises in a specific context of technological progress that makes such an imaging system possible today. Indeed, we can identify three major advances in the recent years that paved the way for functional ultrasound imaging: the increase in electronics performance, the increase in computer power and the advent of contrast agents.
B. Structure and function of the brain
1. The concept of neurovascular coupling
The goal of functional brain imaging is to understand how the brain works. Normal brain activity depends on a continuous supply of oxygen and glucose through cerebral blood flow (CBF). Although cerebral energetic demands are very high, the brain has very little means of energy storage. Therefore, local brain activity has to be matched with a concomitant increase in local CBF, a phenomenon referred to as functional hyperemia or neurovascular coupling. What is particularly striking is that this local increase in blood volume is spatially very confined to the area of the brain activity. Because of this precise colocalization, mapping hyperemia is sufficient to detect very small changes in neuronal activity, such as the cortical column activation by a single mouse whisker, which is less than 100 µm in size.
The concept of neuro-vascular coupling. Credits OBI-Team.
Most functional brain imaging modalities, including fMRI that have provided insight into the human brain at work with unprecedented detail, rely on this coupling. Our technique is based on the same principle.
The key measurements that neurologists want to access with functional sonography are the cerebral blood volume (CBV in mL/100g) and the cerebral blood flow that is the blood supply to the brain in a given time (in ml/100g/min). However, we cannot measure directly these physical units but relative changes in these quantities following activation by a stimulus. We should never forget that these quantities have to be derived from what we can physically measure: the quantity of red blood cells (or contrast agents) moving above a certain speed in the brain vasculature. The hemodynamic response function (HRF) designates the temporal profile of one of this parameter, for example CBV, of brain tissues in response to a correlated stimulus. It is important to note that the vascular response is relatively slow compared to underlying neuronal activity as the time to peak for the hemodynamic response function is around 1s in normal conditions (compared to 300ms for evoked neuronal activity in the same area that can be measured by local field potential with an electrode). Therefore the hemodynamic response can be sampled at a frame rate of typically 2Hz.
2. The architecture of the brain vasculature
The brain vasculature is organized and forms a complex network that drains blood into each part of this vital organ (see Figure below). For example, the cortical vasculature is divided in 3 mains compartments:
- The penetrating arterioles that are perpendicular to the surface of the brain and are directly involved in coupling because they contain smooth muscles that actively dilate or constrict arteries (Example values for rats, diameter from 5 to 30 µm; RBCs velocity from 1 to 35 mm/s).
- The venules that drain the deoxygenated blood outside the brain (diameter from 10 to 50 µm; RBCs velocity from 1 to 25 mm/s).
- The capillary bed forms a dense web of small vessels (diameter < 3 µm; RBC velocity around 1mm/s) that play an important role in the neurovascular coupling although this role is poorly understood.
Vascular architecture of the cortex in rodents. Illustration from Blinder et al., 2013.
Views of the complete vectorized vasculature that show different communities, each labeled by an individual color, as well as columnar boundaries (golden bands).
Most cerebrovascular regulation takes place on the arterial side of the cerebral vasculature, which can be divided into large pial arteries at the surface of the cortex, penetrating arterioles delving deep into the tissue, and capillaries, where most of the oxygen diffusion into the parenchyma occurs. Local CBF changes are induced by constriction or relaxation of smooth muscle cells in arteries and arterioles. Penetrating arterioles are located within regions of neuronal synaptic activity and together with surface arteries, account for a large part of cerebrovascular resistance; hence, these arterioles are most likely the main targets of local neuronal and glial pathways regulating functional hyperemia. This functional network of neurons, glia, and vascular cells is known as the neurovascular unit. In addition, upstream dilation of surface arteries and larger penetrating arterioles is also necessary for adequate and sufficient downstream CBF increases.
II. Functional Ultrasound Imaging – Research & Development
We are currently developing functional ultrasound imaging with a main objective: to exceed all commercially available imaging systems in terms of spatiotemporal resolution and ease of use. This system has been entirely designed by our multidisciplinary team composed of physicists, neuroscientists, high level engineers, biologists and medical doctors and involves many industrial partners.
fUS imaging fits the needs of most of scientists and medics by offering complete compatibility with most standard procedures.
A. A dedicated protocol for brain ultrasound imaging.
Although ultrasound imaging has several advantages, this modality is limited for brain imaging because high-frequency ultrasound waves do not propagate easily through the skull because of attenuation cause by the bone.Hence, this new imaging modality was initially developed in preclinical models (rodents) using a large craniotomy (removal of the entire skull in a specific region) and for an acute experiments where animals were anesthetized during the session and euthanized at the end). A few months ago, our team has developed a new imaging protocol that was inspired from neuroscientists doing intrinsic optical imaging (IOI) or 2-photon (2P) imaging. The main idea is to “thin” the skull over the region of interest, therefore preserving the integrity of the brain and superficial layers instead of removing the entire skull. For the development of the imaging system, we currently use a 1cm2-thinned area in rats from bregma (a landmark feature on the skull used for stereotaxic experiments) +3mm to -7mm AP to record hemodynamics in the somatosensory cortex (including the barrel cortex, the hind and forepaw cortex).
As presented in the Figure 2 (top left panel), we attached an echogenic a bead at the surface of the skull that is used to define the bregma position and we reinforced the surface using a specific optical and acoustically transparent biocompatible polymer resin. This resin is not obligatory but enhances significantly the quality of images during chronic sessions. The figure below (top right panel) presents the immunohistochemistry experiments that validate the innocuity of the surgical protocol as we did not observe any inflammation nor neuronal loss after 2, 10 and up to 30 days after the thin skull surgery. During functional brain imaging, we need to study how CBV and CBF changes as a function of a stimulus and therefore animals have to be in controlled and stable physiological conditions (with limited inflammation).
In the next part, we will present the fUS data performed in head-fixed rodents under anesthesia. As a reminder, our imaging system (Figure 2 bottom panel) is 2D motorized to allow both coronal (vertical plane that divides the brain into anterior and posterior sections) and sagittal (vertical plane which passes from anterior to posterior, dividing the body into right and left halves) of the brain vasculature and therefore is available for 3D acquisitions. During the experiment, body temperature and heart rate are continuously monitored such as the oxygen saturation to adjust the mixture of gas if needed. As we mentioned, one of the main advantages of the brain sonography is the compatibility with EEG recordings that are routinely used by neuroscientists in various experiments for acute or chronic measurements (such as in the case of implanted electrodes).
Presentation of the imaging setup. Credits OBI-Team.
B. Anatomical images of the brain vasculature.
We generated images of the brain vasculature based on micro-Doppler measurement. As presented in the Figure 3, we used tilted plane waves instead of focal beams, to generate compound images that are repeated periodically to produce a video of brain echoes. As explained below, given the blood velocity distribution in rodent brain vasculature, we selected an image frequency between 200-500Hz.
We will now briefly describe here a part of our results based on plane wave imaging. High quality anatomical images of brain vasculature are composed of 15 plane waves between -15° and 15° (taken every 2°) that are emitted at a pulse repetition frequency (PRF) of 7 kHz. This block of 15 plane waves is repeated 200 times to give a compound image in 400ms (2 cardiac cycles in rodents) at a 500Hz frame rate. Ideally, the PRF should be the highest possible for a given imaging depth (typically 20 kHz for the rat brain) but it is currently limited by the scanner electronics.
Next, all data are transferred to the computer with a single transfer-to-host instruction at the end of the sequence. As shown in Figure 3 (bottom right panel), if you integrate the signal in each pixel over the time, the resulting micro-Doppler image displays intensity proportional to the number of red blood cells that are moving inside each voxel during this period of time; thus, it traces out the position of the blood vessels. The signal processing also includes a set of intelligent data filtering schemes for suppressing tissue motion.
The micro-Doppler images of the brain vasculature that we produce are so resolved that we can observe all the penetrating arterioles and venules in the cortex, a less perfused region corresponding to the white matter, and subcortical regions (hippocampus, thalamus) containing vessels oriented in various directions.
Anatomical images of brain vasculature using plane wave compound imaging. Credits OBI-Team.
The quality of these vascular images obtained in chronic conditions in only half a second is unprecedented, using ultrasound imaging or any other imaging modality. This result is extremely motivating for us as it demonstrates that brain sonography can rapidly become an essential tool for neuroscientists.
C. Functional imaging of brain activity: Sensory evoked activity.
What is currently known about brain function mainly comes from studies in which a task or a stimulus is presented to the subject and the resulting changes in neuronal activity or behavior are measured. This approach to study the brain function has been very successful, from the pioneering electrophysiology work to cognitive activation paradigms in human neuroimaging (giving insight on reading, learning, etc). A simple example of this approach is a paradigm that requires subjects to open and close their eyes at fixed intervals. Modulation of the functional hemodynamic signals attributable to the experimental paradigm can be observed in distinct brain regions, such as the visual cortex, allowing one to relate brain topography to function.
As presented in previous chapters, we can indirectly record neuronal activity of the brain using neurovascular coupling. We also showed that we can acquire vascular images, sort of a sort of "snapshot" of blood volume distribution in the brain at a specific time. Functional imaging consists on acquiring such images continuously during the stimulation paradigm to follow changes in blood volume triggered by the stimuli.
Unfortunately, the buffer of the imaging system is limited. To perform periodic acquisition, the repetition of the vascular sequence can be controlled by the software using an end-of-sequence event or directly with the hardware by separating the memory into two sequential buffers played in loop (one used for recording, while the other transfers data). We chose this last method that allows for continuous acquisition of the hemodynamic variations - if the transfer time is shorter than the acquisition time.
The top right panel in Figure below shows a representation of the field of view when using a linear array oriented in a coronal plane above the primary somatosensory forelimb cortex. Contrary to classical optical techniques such as intrinsic optical imaging (IOI) commonly used in preclinical studies, the penetration of ultrasound is adapted to see hemodynamics through the entire brain including subcortical relay centers such as the thalamus or the basal ganglia.
To validate our new modality, we performed various evoked stimulations, such as presentation of a non-noxious electrical stimulation of the hind paw, the forepaw or the whisker pad. As presented in the Figure below (bottom left panel), we used a stimulus paradigm that is composed of a 20s short block containing a 5s baseline, then a 5s stimulus (5Hz current pulses) and a 10s post-stimulus period. We recorded one compound image every 0.5s that are processed with a short delay (computation time) by the computer. We developed 2 algorithms for functional imaging that will generate an activation map to show the location of activated voxels (z-scores) or a correlation map to localize the increase in hemodynamic signals that are correlated with the stimulus. The figure below (bottom right panel) shows a clear correlation in the cortical area, called forelimb cortex, after the stimulation of the left forepaw as expected because brain wiring is in many cases contralateral. As observed here, we measured an increase of up to 20% in CBV in only one trial.
Functional ultrasound in anesthetized rats. Credits OBI-Team.
A. Schematic representation of the head of the rat showing the localization of the different areas that could be activated (S1 primary somatosensory cortex, FL: fore limbs, HL: hind limbs, BF: barrel field). B. Position of the transducer during this experiment. C. Experimental paradigm. D. functional ultrasound image showing the activated area (red) in the cortex after a single stimulation of the left forepaw.
This result illustrates the main advantage of our new imaging modality compared to fMRI or optical imaging. Indeed, there is no need to average in space or time because this technique offers an unprecedented signal-to-noise ratio (SNR) and exceptional spatial resolution. We recently demonstrated that a single trial is enough to visualize functional activation in chronic conditions.
D. Functional ultrasound imaging in real-time.
To enhance performance of the imaging system, we are developing a dedicated HPC workstation and GP-GPU beamforming and filtering which paves the way to real-time functional imaging.
We are currently evaluating others high power computing architecture such as Massively Parallel Processing Array (MPPA) or parallel reconfigurable architecture combining multiple FPGAs.