GUILFORD, Conn. — In a nondescript building in this seaside town sits a prototype that could change the way that doctors use magnetic resonance imaging.

Usually an MRI machine requires a giant, powerful magnet and must be encased in its own copper-shielded room. It is why the behemoths cost hundreds of thousands or even millions of dollars. But this device, from a startup called Hyperfine, is about the size of a luggage cart. It could be wheeled from one bed to another.

On a recent day, I lay down and put my head under the machine. Within minutes a three-dimensional image of my brain was on a screen in an adjoining room — not quite as good as a normal MRI, but pretty clear. There was no need for me to take off ordinary metal objects. I wore my belt.


“If somebody is really sick, they can’t be moved in a hospital. So we can look at that person,” Jonathan Rothberg, the serial entrepreneur who started Hyperfine, said in an interview. “If you want to see changes over time, you can’t do that with a classic MRI machine. We can be in the room. We can see it over time. And we have to have a good enough image to be actionable.”

In addition to being small enough to be wheeled around or even put in a van, it will be significantly cheaper than existing MRIs: The price will be about $50,000, Rothberg promises.

It’s impossible to judge such a technology just by seeing it once. But Rothberg has a knack for radical thinking, and has tried to push medicine to be more like the semiconductor industry — in which things get less expensive over time, not more. His previous companies were leaders in DNA sequencing and were sold for hundreds of millions of dollars to larger life science firms. Two years ago, another of his companies, Butterfly Network, unveiled a handheld ultrasound that hooks up to an iPhone and, at $2,000, costs just a fraction the amount of other devices. Now he’s trying to replicate that model with Hyperfine in MRI.

Jonathan Rothberg
Jonathan Rothberg, the founder of Hyperfine Matthew Herper

Dr. Kevin Sheth, the associate chair for clinical research in the department of neurology at Yale University, said the device has been used more than 100 times in the neurological intensive care unit he oversees as part of ongoing clinical research. He has not personally received any money from Hyperfine.

“For some of the sequences, the pictures have gotten quite good,” Sheth said. “For others, they’ve gotten better but still have a ways to go.” Are the pictures useful enough for making clinical decisions right now? Maybe not, Sheth said, but he thinks they will be soon. And he pointed out that for some uses, like seeing whether brain swelling or a bleed is getting better or worse, a device like the Hyperfine model could be advantageous. It can be moved to the bedside of a patient, whereas a traditional MRI cannot.

Hyperfine has filed with the Food and Drug Administration for its first clearance, for using the device for imaging the brains of adults and children over 2 years old. And it is working on using the device to image patients’ feet, ankles, knees, and other body parts.

Rothberg’s past efforts have generated obvious and conspicuous wealth. In the late 1990s, he founded Curagen, a genetics company that soared amid the hype around the sequencing of the human genome. It eventually crashed, and was sold to a small company, Celldex, for just $100 million. But by that time Rothberg had moved on to start another firm, 454 Life Sciences. It was the first to make it faster and far cheaper to sequence DNA — the company dropped the cost of sequencing a human genome by 100-fold to $1 million, and garnered attention by making the first genome it sequenced that of Jim Watson, co-discoverer of DNA’s double helix structure.

But 454, which was sold to Roche (RHHBY) for $155 million in 2007, was eclipsed by Illumina, still the dominant DNA sequencing firm. Rothberg made a second go at the DNA sequencing market with another company, later sold for $725 million, and he left within a few years.

These days, Rothberg runs a startup incubator, 4Catalyzer, that stretches across a campus of multiple buildings in Guilford. Here, he has started seven companies, of which Hyperfine is the second to come out of stealth mode.

Guilford is a small town; the building where Rothberg keeps his MRI is also home to the maker of sailboat parts. He sometimes eats at a small food stand, called Pa’s Place, named for his dad. He also has his own sandwich, Jonathan’s Tuna, that consists of seared tuna in wasabi mayo with cucumbers on a kaiser roll.

Rothberg became interested in MRIs 20 years ago because of claims that, with the right contrast agents, it might be possible to use them to see genes turning on and off. That turned out to be impractical at best. But several years ago, he took a group of experts out for a spin on his yacht, the Gene Machine, out of Fort Lauderdale, Fla. (He now has a new yacht, with the same name.) He joked with them that he was paying the captain, and that they might not return until they had sketched out a solution to a problem: How could you make an MRI machine that could be built for just $10,000, that wouldn’t interact with metal objects, that would not need its own room, and that would be just a little bit portable?

“When MRI was launched,” Rothberg said, “computing was 10 million times more expensive. Ten million times.” The brainstorming session on his boat led him to think that he could combine better computer power with advances in magnets that had come out of the green energy industry (think windmills) to make a bedside MRI device.

The trick from the start was to use all that computing power to get a useable image out of a much smaller magnet. An MRI works by using a magnetic field to align protons, disrupting them with a radio frequency, and measuring the energy they release. The solution to making better MRIs has meant bigger and bigger magnets.

The strength of an MRI’s magnetic field is measured in a unit called a Tesla. Most MRIs are 1.5 Teslas; some higher-end models use a 3 Tesla magnet. Before he joined Hyperfine, the company’s principal engineer, Michael Poole, was working in Germany on getting higher quality images from a 9.4 Tesla system. The Hyperfine device is built around using a magnet that is just 0.064 Tesla. “It’s quite a shift,” Poole said.

Dr. John Martin, a vascular surgeon who joined Rothberg’s group to work on ultrasound and is now also overseeing the MRI work, said the whole concept of the Hyperfine device takes what doctors think they know and turns it upside down. “Now, the cool thing is that when you get this in front of doctors, they look at it, they start thinking about all kinds of creative ways in which they can use that,” he said. “You know, the thought of this now being in a van floating around Africa, that’s a mind-blowing concept, no one would ever dream of that at all. That’s actually possible.”

Hyperfine says the unit is 20 times less costly to build, 10 times lighter, and uses 35 times less power than a 1.5 Tesla MRI. And it’s spurring Rothberg to dream big.

Walking around a garden he built at 4Catalyzer, wearing a sweatshirt adorned with drawings from the artist Keith Haring that one of his daughters bought for him, he imagines a big future for his new machine. He talks about how some doctors have used the device to image infants’ heads. His ambition, he said, is for the machine to lead to an entirely new kind of paradigm, and to make sure his team is using deep learning to make its new product continually better.

“We’re definitely inventing something,” he said. “Point-of-care MRI doesn’t exist. A portable point-of-care MRI doesn’t exist. And it’s great to fork history, but it’s more important to own that fork. So I hope we take enough lessons to do so.”

Leave a Comment

Please enter your name.
Please enter a comment.

  • The possibility of a portable low field magnet usable vs. a high field fixed magnet is exciting but also frightening for the reasons the unnamed ‘Radiologist’ gives. If the charges for a 7T or a .01T magnet are similar, the patient will be poorly served. However, if the expectations AND the charges for the very low field magnet are in-line, this may be a very beneficial technology which spurs faster diagnoses and opens up treatment options that are currently unavailable. I am speaking specifically about stroke treatment, with the unit usable in an ambulance designed specifically for rapid response. In larger urban centers, this might be a significant game changer reducing disability and long term care costs.

    But let’s not put the cart before the horse – such a unit would need to prove its worth. And the sequences that stroke imaging is dependent upon are among the more difficult in terms of physics. Deep learning is capable of exciting feats, but whether it is able to solve the S/N ratio problem with low field imaging is another story. For to do so, it must ‘hallucinate’ the missing data – can it do so without hallucinating either pathology or normalcy? Big, big, ifs, and as a radiologist who experiments with deep learning, I speak of this from some experience.

  • Rothberg’s idea and intentions are laudable, and he puts his substantial energy to work for medical improvements. However I tend to agree with the Radiologist comment, that only high definition images are useful diagnostics. To sacrifice result for convenience / access is a backwards move, and that needs to be ironed out first. “Preliminary MRI’s” would only run up costs.

  • The real benefit is in developing countries for both ultrasound and MRI. However this will facilitate gender choice in babies resulting in more abortions of females in utero.

  • Please don’t do this – it is a step in the wrong direction. Going toward low field magnets that do a bad job just because they are cheaper and more convenient than modern 1.5 and 3T magnets is bad for patient care. There is currently very little clarity for the consumer in terms of what kind of scan they are buying. They could go for a wrist mri in a state of the art 3T and get excellent images that accurately diagnose their problem or they can go to a low field magnet with techs that do not know how to scan a wrist and end up misdiagnosed. They still get the same bill. It isn’t fair. You don’t go buy a car and have the sales person say pay first and then you can find out what kind of car you bought. We should not be doing this in medicine either. If your low field convenient scanner is to come into the market, you should be paid much much much less for its use than a modern 3T and should have to clearly explain to the patient that your machine is substandard as are the images it produces.