Could brain imaging become as routine as ultrasound scans?

Where MRI and CT scanners take up entire rooms and are expensive, ultrasound is a much cheaper and smaller alternative. But doctors don’t typically image the brain with ultrasound, as it can’t easily penetrate the skull and the images it provides are inaccurate and distorted.

That could be about to change. An Imperial College London spinout, Sonalis, is repurposing algorithms originally invented for oil and gas exploration, along with sensors from nondestructive testing, to boost medical ultrasound far beyond what it’s currently capable of. The hope is to make brain imaging much more accessible to hospitals with smaller budgets, for example in low-income countries.

Oscar Calderon Agudo and two other seismologists at Imperial founded the company in 2017 after seeing parallels between visualising oil and gas reserves and visualising the brain. 

Between hardware development, attracting investment and gaining acceptance as outsiders among the neurotech research community, the journey was at first an uphill struggle. “It’s taken a long time to make these theoretical predictions and simulations a reality,” says Agudo. Nearly eight years later, the company has proven its imaging technology in human testing. 

Going underground

Identifying underground reserves of oil and gas is a tricky problem to crack. Oil and gas companies have pumped “millions and millions” into developing seismic imaging that reveals complex underground geology, says Agudo. 

By sending low-frequency sound waves into the ground, capturing reflected energy and analysing the signals with specialised algorithms, geophysicists can discern the presence of oil and gas reserves, such as beneath the bodies of salt commonly found near the reserves. As it turns out, the skull’s capacity to reflect waves is not unlike that of these underground salt bodies.

In a typical medical ultrasound scan, a handheld scanner sends ultrasound waves into the tissue. Weak reflections from boundaries between different tissues are used to produce the resulting image, but with the head, most of the signal is reflected from the skull surface.

Agudo and his co-founders wondered what would happen if they surrounded the head with sensors and analysed both the reflected and transmitted energy. By coupling this data with the algorithms applied in subsurface exploration, they could model the way the waves propagate through bone and tissue. 

The prototype device resembles a helmet packed with hundreds of sensors. This first helmet cost a fraction of the cost of a commercial MRI machine, and with optimisation, the company expects the cost will fall further.

At first, the images were “very noisy”, says Agudo, comparing them to CT scans from 30 years ago.

But one day in 2025, the cortical folds – the characteristic ridges and grooves on the surface of the brain – and central cavities, known as ventricles, came into view. These areas of the brain are where clinicians look when a patient is suspected to have had a stroke or brain haemorrhage.

It was a thrilling moment. “We were like, oh wow, that’s amazing. Nobody in the world had seen this before.” 

Driven by the AI boom 

Sonalis’ brain imaging technology has come to fruition partially thanks to the exponential rise in computing power driven by the AI boom. “Now, it’s feasible in a very short amount of time [to produce] reconstructions of the brain. That was not possible two decades ago,” says Agudo.

New sensors have been essential, too. Existing medical ultrasound usually operate at high frequencies, which are highly absorbed by the skull. Instead, Sonalis employs lower frequencies, which are safer and penetrate deeper into the brain. The team sourced sensors from nondestructive testing of pipes, where the sensors can pick up ultrasound frequencies a thousand times lower.

You might expect with lower frequencies, and thus lower energy, that the images would be lower resolution. However, thanks to the algorithms, they are approaching millimetre-scale resolution.

This precision is limited by current sensors and has not yet peaked, the company believes. “As the technology evolves, this resolution will get better and better. It’s just a matter of time,” says Agudo.

A path to treatment

Imaging the brain with ultrasound is just one of Sonalis’ goals. The spinout also aims to treat neurological conditions such as epilepsy with its technology. 

The idea is to focus the ultrasound at a precise point in the brain, concentrating the ions, and combine it with an electric field in this small region to stimulate the neurons. 

Currently, stimulating neurons can only be done by very invasive methods, such as inserting electrodes in the brain after removing parts of the skull – so the prospect of doing this with no surgery required is exciting. But it’s still early days, stresses Agudo. Further investigation is needed to explore suitability for epilepsy and other conditions, such as Parkinson’s and drug-resistant depression. 

Moonshot research

The journey has not been without challenges. With the founding team’s home territory in geophysics and seismology, publishers and funders were initially reticent. “There was a bit of scepticism at the beginning and we kind of persevered until we got published,” says Agudo. 

Another difficulty thrown up by operating in speculative, uncharted territory: off-the-shelf hardware wasn’t an option. “We were a moonshot project,” says Agudo. “We had to build our own labs to build this hardware.” 

And as many startups find, there was a slow grind to accumulating enough funding. But over time, the team amassed enough grants to build their proof-of-concept device, and since 2023, things have taken off. 

Investment from the UK’s Advanced Research + Invention Agency (ARIA), has accelerated Sonalis’ trajectory, explains Agudo. As well as ARIA’s funding and support in navigating regulatory bodies, Sonalis could tap into a unique network of engineers that helped them design parts of the current device. 

At the current rate of development, Agudo estimates the device will be in clinical use in the next five or six years. But with enough of a boost from an ongoing fundraising round, this could be cut down to two and a half years, and as planned, the technology deployed in low-income countries. 

“Where people have struggled to access to this high-resolution technology, [it will] benefit as many people as possible,” says Agudo. “That’s the goal.”