We’ve all seen ultrasound images of foetuses in their mother’s womb. But using these particular waves to “dig” deep into the inner workings of the living brain is virtually unheard of. It may not be for long.
Alan Urban is a 39 year-old French scientist who is principal investigator at Neuro-Electronics Research Flanders (NERF) – a joint research initiative between IMEC (Inter-University Micro-Electronic Center), VIB (Flanders Institute for Biotechnology) and KU Leuven (Katholieke Universiteit Leuven) –, in Belgium.
For almost a decade, his professional goal has been to build machines based on good, old ultrasound waves, that will allow doctors to visualize the functioning brain of patients in all its depth. He was trained as a molecular and cellular biologist – “physics came later”, he says.
He explains why he firmly believes that the ultrasound imaging technology he is contributing to develop could become essential, in conjunction with other imaging technologies such as magnetic resonance, for the study of the functional – and dysfunctional – brain, both in research and in the clinic. A few weeks ago, he gave a talk on the subject at the Champalimaud Centre for the Unknown.
You specialize in functional ultrasound imaging of the brain. Can you explain what that is?
Functional ultrasound imaging (fUSi) is a technique that aims to understand how the brain works by imaging brain function – that is, by looking at how cerebral circuits are activated when an animal is performing a motor task, be it a memory task or any sort of somatosensory task that involves feeling, touching, hearing.
Just as fMRI (functional magnetic resonance imaging), fUSi looks at brain activity at the global scale. We look at blood variations instead of induced magnetic fields. There is a very strong link between brain (neural) activity and blood activity because, when a neuron starts sending electrical information, it needs more energy.
That’s why, very rapidly, in the region around the activated neuron, there’s going to be an increase in blood volume, which happens mainly through an increase of the flow of blood. Functional ultrasound imaging analyses the Doppler signal associated to the movement of red blood cells.
What is the spatial resolution of the technique?
Let’s do some physics. We know it is possible to image an object using a physical phenomenon whose wavelength matches the size of the object we want to observe. At the frequency of the ultrasound waves we use, we will not see the smallest blood vessels, the capillaries. Nonetheless, when we increase the wave’s frequency, we can go down to a resolution of 100 micrometers, which is just 10 times the size of a capillary or of a neuron. This is the size of the so-called penetrating arteries, which are the blood vessels that receive blood from the body’s main arteries and bring it into the brain, then further branching into smaller capillaries.
But you cannot visualize the fine structure of the brain’s vascular system…
As I said, we know we don’t have the necessary resolution for that. But we can detect overall signal variations inside what we call a voxel – a unit of volume – as a function of time.
To give an analogy, the naked eye cannot see a car’s headlights from five kilometers away, so we cannot tell exactly how many cars we are looking at from such a distance. But what we can tell is whether headlights are on or off.
That’s the basic idea here: we can’t tell just how many capillaries there are, but if the signal is strong enough and the contrast (signal to noise ratio) good enough, we don’t need to see the individual capillaries. What’s important is that in each unit of volume, we will detect anything that lights up.
So the technique can pinpoint where things are happening in our images – and in the brain. Not down to a tee (capillaries are 10 times thinner than our resolution), but we are highly confident in our capacity to detect very fine variations of activity.
Who developed fUSI?
It was initially developed in 2011 by two physicists at the École Supérieure de Physique et de Chimie Industrielles (ESPCI), near Paris: my colleague Gabriel Montaldo and his then PhD student Émilie Macé. Gabriel is now a senior researcher in my team. We wanted mainly to improve the technique for neuroscience applications.
In the beginning, we applied it to animals that were anesthetized, with their head in a fixed position, which is the simplest configuration. Now, we are working with animals that can move freely either in a confined or open space and perform behavioral tasks, and see what’s going on in their brain when they are feeling or seeing something. Ultimately, we want to use this tool also in humans.
How did the idea to use ultrasound in this way arise?
As a biologist, I realized there were very few tools to detect brain activity in real time during long periods of time. I was at ESPCI at the time, and I met these two physicists, who wanted the same thing but didn’t necessarily have my biological know-how. It was a snug fit, and we decided to do it together.
The simple idea was that, because of their wavelength, ultrasounds are the perfect way to answer questions about the human body.
Communications use radio waves. Cancer radiotherapy uses X-rays to look at samples. We use light to see. But the best wave with which to look at the human body are undoubtedly ultrasounds. They match organ size, and when we increase the frequency we can go down just an order of magnitude larger than neurons. It’s also the ideal wave to go into the brain because of its penetration.
What distingues your ultrasound technique from standard ultrasound?
The whole difference resides in the new imaging mode that was developed by my colleague Gabriel Montaldo. There are two important elements: the first is that instead of using focused ultrasounds to scan the brain, we use so-called flat waves. So the way we generate the waves is different from standard ultrasound, and this requires more sophisticated electronics. The other important element is the extremely sophisticated data post-processing and analysis we perform.
There is also a third difference: the relatively high ultrasound frequency we use. Contrary to standard medical ultrasounds, which are between one and ten MHz, we use ultrasounds at 15 to 30 MHz. The higher the frequency, the higher the resolution (but this will lower penetration of the wave, so we have to tweak these parameters to find the optimal conditions).
What about temporal resolution?
The temporal resolution is a bit “on demand”. The imaging frequency we recommend in small animals, because we saw it gives very good results, is around 10 images per second [one image equals one full ultrasound scan of the brain]. This is sufficient to detect very weak activations in the brain.
Is this a real-time imaging technique, just like standard ultrasound?
Ultrasound imaging was developed as a real-time technique. When performing pediatric ultrasound imaging [in pregnant women] – which is not functional, but rather structural – to look at a baby, we don’t need to wait: the image appears instantly. This allows doctors to see the baby move, see its tiny feet, it can even be done in 3D.
We wanted to preserve this capacity to display an image as fast as it takes to acquire the data. That’s what we call real time: if you need 100 milliseconds to acquire all the data for an image, you have to be able to display it in 100 milliseconds. To the human eye, the image seems to appear as it is being recorded.
How do you process the images?
We subtract a baseline image from the images obtained during brain activation, and analyze the zones where an increase in the amount of blood took place. We obtain a timeline of when the voxels became activated, which tells us where neural activity was taking place as function of time.
What are the advantages of fUSi with respect to fMRI?
MRI machines are huge magnets and impose huge constraints. Because they have to be kept from moving, in 90 percent of the cases the animals are anesthetized – so they cannot perform any task. And if they are awake, they have to be kept absolutely still.
With fUSI, we can do imaging on animals that are free to move on a surface the size of a desk (1-2 square meters) – or inside mazes to perform memory tasks. In fact, fUSI can even be performed on a “chronic” regime, keeping the ultrasound device implanted in the animal for days, months or even years.
A window into the brain
Ultrasounds do not penetrate bone, so you would literally have to open a window in the skull of a patient to scan his or her brain. Isn’t this a big limitation?
It is precisely one of the advantages of MRI with respect to ultrasounds, which are strongly attenuated and distorted by hard tissues such as bone. Computationally correcting these aberrations is also not possible, because no two brains are identical.
However, some years ago scientists developed ways to perform very stable skull bone replacement in animals, where the ensuing inflammation can be very well controlled. The bone is replaced by a plastic polymer and the animal acquires a skull that is transparent to ultrasounds.
Right now, with the condition that in the adult human brain you would have to remove a piece of bone, we obtain penetrations of several centimeters; with a wave at eight to ten MHz, the penetration will be around five to six centimeters, which is more than enough to perform most of the standard medical exams that are needed. So we manage to have a very good field of vision. In the case of infants, we can see the whole depth of the brain without removing any bone.
In humans, the most obvious applications, in my opinion, are precisely in newborns, because their skull is not yet completely closed – especially the frontal fontanelle. The fontanelle, which is made of cartilaginous tissue, is a natural window into the baby’s brain during a few months after birth. This allows brain imaging in a completely non-invasive way, and could be used for brain function assessment in cases of neonatal hypoxia, where the baby’s brain has been deprived of blood. Diagnostic purposes are one of the things we want to use fUSI for.
How could it be used?
When a child is born, a series of functional tests is performed, to check if it cries, if it reacts normally to light. But sometimes babies are hypoxic, they barely react, their skin color indicates poor blood circulation, they are very passive. They need to be diagnosed quickly, to know if they can benefit from so-called brain-cooling, or body-cooling, which consists in lowering body temperature by a few degrees for a few days, hoping that the brain will self-repair to a certain extent. There’s not much else to be done today.
We hope that our technique will allow a better qualitative understanding of the state of the baby at the start. Is the brain damaged, and in that case, is the damage at the level of the visual system, or brain-wide, or the somatosensory system? All hypoxia cases are different, each baby has its particular problems. Currently, none of the standard tests are based directly on cerebral activity; they are indirect: you tap the baby, you talk to it, touch it, you pinch it. We firmly believe that if cerebral activity can be imaged, that will be the most important indicator of the baby’s state.
Will it be like performing a conventional ultrasound?
For this application, we developed a special device to be fitted on the baby’s head, a sort of a helmet to which we apply the ultrasound probe. The baby will then be stimulated while brain activity is recorded. At the location where the probe is positioned, we can see the whole depth of the brain. The field of view of the probe is around two to three centimeters wide and seven to eight centimeters deep. This is relatively unique, and allows us to cover most of the functional areas of interest.
Is your system already in use in the clinic?
Not yet. A clinical trial will begin in Belgium in a few months to test the feasibility of this application. We collaborate with very well-known neonatologists in the area of ultrasound cerebral imaging. Instead of performing a series of tests that take ten minutes and which give a score that doesn’t tell you anything, we want to obtain images that provide real information.
This will take some time, several years. We also want to do prognostics, that is to try to predict, after a few days of body-cooling, what will be the long-term impact of hypoxia. Impairments, paralysis may ensue, leaving the child handicapped for life.
There are also animal models of human diseases, in particular in mice. Will fUSI it be useful to study brain function in these animals?
It’s a very interesting question. In fact, a huge number of serious brain pathologies are associated to an impairment of vascular function in the brain. It’s not big news, but it shows that the heart-brain axis is extremely important. If brain arteries age and harden, this increases the risk of aneurysms and stroke. Neuron health depends on the quality of vascular function – and when neurons die, it is also because the tubes that bring oxygen into the brain are damaged.
More and more, people are realising that neurodegenerative diseases are in fact vascular pathologies. Recent study have for instance shown that in animal models of Alzheimer’s disease, if you clean up the vascular system, you can restore memory function in mice that had lost it – that is, in animals with Alzheimer-like impairments. So maybe it would be possible, with a brain imaging tool like ours, to detect vascular dysfunction in the brain, in order to diagnose this type of pathology at an earlier stage.
In the case of stroke, there are not many preclinical tools for mapping the vascular system. In humans, it is unlikely fUSI will be used because of the need to remove the skull. But in small animal models there is still a lot of research on stroke that has to be done. This tool makes it possible to map the amount of blood in the brain very precisely and in real time. We can see the blood volume across the whole brain in less than 20 seconds.
Ana Gerschenfeld works as a Science Writer at the Science Communication Office at the Champalimaud Neuroscience Programme
Photo: Tor Stensola.