Canadian team is making ultrasound easier to use
March 31, 2023
An Edmonton-based team that is part of Exo has produced AI-based algorithms that enable healthcare providers with minimal training in sonography to take skillful ultrasound scans of the thyroid gland and to detect dysplasia in infants.
The team was created a few years ago as its own company, Medo.ai, but its expertise recently attracted the attention of California-based Exo (pronounced Echo), an ultrasound technology company, and it purchased Medo last year. While it’s now part of Exo, the Edmonton team is staying in place.
“We’re located right in a building with AMII (the Alberta Machine Intelligence Institute),” said vice president of AI at Exo and former Medo.ai co-founder, Dornoosh Zonoobi, PhD. In fact, the group is hiring more employees, adding to its team of 14 in Canada.
Both the hip dysplasia algorithm and the thyroid ultrasound algorithm, to aid qualified users, have gained FDA clearance in the United States.
Dr. Zonoobi explained that her team’s AI technology is solving some major problems. While ultrasound machines are highly portable, the expertise needed to take and interpret scans is often unavailable in remote communities.
As a result, patients needing exams in remote areas are often transported at great cost to an urban centre. In non-urgent cases, the patient must find their own way, causing them lost time and added expense.
With the solution, created by Dr. Zonoobi’s group and to be commercialized by Exo, caregivers with even minimal training in ultrasound can take accurate exams.
“We have nurses who have never used ultrasound before using probes to check babies for dysplasia,” she said.
This can have an enormous impact on the lives of individuals. Dysplasia of the hip – when the femoral bone does not fit perfectly into the hip socket – will lead to pain and suffering in later life, and most likely to a hip replacement by the time the patient is in their 40s.
Doctors check for it at birth, but they tend to miss 80 percent of the cases when only conducting a physical exam.
An ultrasound exam has far higher sensitivity for detecting dysplasia, said Dr. Zonoobi, but most birthing centres can’t afford to send all of their babies to radiology departments for the exam.
With Exo, on the other hand, the exam can be done right at the point-of-care.
Exo is testing the technology at sites across Alberta, said Dr. Zonoobi. After checking 300 newborns, they detected dysplasia in six babies, and the researchers will be writing a paper about their results. Exo is hoping that word will spread, and that the solution will be adopted across Canada and beyond.
On a related note, she observed that Indigenous communities tend to have a much higher incidence of dysplasia in newborns – up to 30 times higher than the general population – and the majority of cases are missed.
“It’s a driver of opiate use in later life,” said Dr. Zonoobi. “The patients face a great deal of pain.”
Exo is currently working to add more sites for the detection of dysplasia using its ultrasound solution. “We’re aiming to screen 20 percent of newborns in Alberta in the next phase,” she said.
The company has also devised an impressive AI algorithm for ultrasound exams of the thyroid gland.
Dr. Zonoobi explained that thyroid exams are lengthy – they typically take 30 to 40 minutes of an ultrasonographer’s time. They’re also highly prone to error, as the sonographer must be skilled in the art of properly sweeping both sides of the gland.
With Exo’s AI-powered system, however, any technologist or care-giver can perform the exam and the artificial intelligence finds and fills in the needed information.
Moreover, this can be done in a fraction of the time needed for a traditional thyroid ultrasound examination. “Instead of 30 to 40 minutes, it takes five to 10 minutes,” said Dr. Zonoobi.
The exam can then be sent to a radiologist for interpretation.
“We’re taking away the variation in exams that occurs with different sonographers,” she said. And by cutting the exam time, more patients can be seen each day.
In the United States, for example, over 1.5 million thyroid ultrasound exams are done each year. Lumps or “nodules” often appear on the thyroid gland, and while usually benign, they can become cancerous in some cases and require regular check-ups.
In addition to reducing exam time, Exo’s solution also assists radiologists by selecting the optimal images for analysis, calculating measurements (a tedious task for the radiologist), and characterizes any nodules present using TI-RADS – short for Thyroid Information-Reporting and Data System.
Further, the system contains several breakthroughs, including a cross-referencing ability previously only possible on multi-plane CT and MRI. This feature assists the radiologist with viewing nodules across all planes of interest, such as transverse and sagittal views.
She said that when it comes to AI, Exo’s solution is not a ‘black box’, referring to the phenomenon of an algorithm performing work but end users not knowing how it did it. “Everything is verifiable,” she said. “You have to allow the user to verify the results.”
Being able to do this and seeing that the solution provides accurate results over time leads to physicians and other professionals gaining trust in the AI solution, she observed.
Dr. Zonoobi said her team at Exo is now working on additional types of ultrasound exams and that announcements will be made in the near future.
She commented that eventually, with the help of AI, ultrasound exams will be able to be taken in the home by consumers. Untrained users will be able to take accurate exams and the results can be interpreted by the algorithm or by sending them to a radiologist.
To this end, Exo in the United States is working on a highly portable, point-of-care ultrasound device, a powerful but tiny instrument that will be available for use in hospitals and clinics. It’s currently a work-in-progress, but it’s one of the company’s major goals. “It will be like the Tricorder in Star Trek,” observed Dr. Zonoobi. “Exo is building it, plus a whole ecosystem of applications.”