Hospitals use Imagia’s AI platform to find answers to clinical questions

MONTREAL – Imagia, one of Canada’s leading AI companies, announced that six hospitals in Canada and the United States are now using its Imagia EVIDENS platform to apply machine learning to healthcare and to create better therapies for patients.

The hospitals now signed on to use the EVIDENS platform are:

• Centre hospitalier universitaire de Québec – Université Laval
• CIUSSS de l’Estrie – Centre hospitalier universitaire de Sherbrooke,
• University of Montreal Hospital Centre (CHUM),
• McGill University Health Centre (MUHC) and its Research Institute (RI-MUHC),
• Jewish General Hospital (JGH), and
• Penn Medicine.

Together, the hospitals represent a pool of more than 4 million patients treated annually. It’s important for AI researchers to have access to large amounts of data, as a larger data set is believed to be the key to achieving accurate results in personalized medicine.

Already, three of the hospitals have embarked on a collaborative project to improve medical outcomes for lung cancer patients by analyzing treatments and results with the EVIDENS platform.

The partners collaborating on the lung cancer project are CHUM, McGill and the Jewish General. They’re studying tests, therapies and outcomes to find the best treatment plans for these patients, from mainstream chemotherapy to immunotherapy.

Florent Chandelier, CTO at Imagia, observed that the EVIDENS platform is enabling “federated learning,” with researchers and clinicians at each hospital learning from one another.

An important feature of the “federated” system is that it allows users to share the AI results and insights of the work being analyzed and processed without divulging patient identifiers or raw patient data, as these remain inside of each hospital. That’s important for patient privacy; it enables researchers to collaborate effectively while respecting patient privacy rules across Canada and the United States.

It’s not difficult for researchers working in different languages to use EVIDENS, as the system is designed for multi-lingual usage. Moreover, Imagia has strengths in natural language processing, Chandelier said, allowing the system to convert healthcare data into information that is searchable in a language of the user’s choice.

He noted that researchers in each of the hospitals will be using EVIDENS with their own data and local projects. At the researchers’ discretion, they are free to collaborate with other users of EVIDENS in federated learning.

“The principal investigators can invite other members,” said Chandelier. “It’s in the hands of the researchers to scale their discoveries across institutions.”

For its part, Imagia will acquire any intellectual property that’s created in the projects, and in fact, the company hopes to commercialize some of the solutions. Chandelier said useful results in personalized lung cancer treatment are expected as soon as this year.

Additional hospital partners are to be added, said Chandelier: “This is just the start.”

Dr. Michaël Chassé, scientific director of the CHUM Centre for Integration and Analysis of Medical Data, said the EVIDENS platform is enabling researchers at different hospitals to work in a secure way. “We’re very conscious and careful of patient privacy,” he said. “We want to make sure that all laws are respected. We can do this using EVIDENS.”

He said connections have been built linking data at the three hospitals working on the lung oncology project. Pooling real-world evidence derived from each institution’s healthcare data is achieved through federated learning.

This allows researchers to access and learn from a wide variety of data without patient identification ever being disclosed. It includes all manner of radiological images and reports (e.g., X-ray, CT and MRI), lab data and pathology reports, as well as other information.

Dr. Chassé noted that pathology images themselves are just starting to be digitized, but the text reports are accessible to the researchers.

All of the AI discoveries can be shared using the federated system, which Dr. Chassé said is a fairly new technology.

And while help is available to researchers from Imagia when they need it, the company says its EVIDENS platform is designed to be user-friendly. “We are the first to combine automated AI model discovery and federated learning in a hospital setting for researchers to undertake different discovery processes and validate clinical intuitions, without requiring a team of AI experts by their side,” said Chandelier.

Imagia was launched in 2015 by long time friends Alexandre Le Bouthillier and Nicolas Chapados. Yoshua Bengio, known as the “godfather” of Deep Learning, joined them early on as Imagia’s scientific advisor.

Three years later, Imagia acquired Chandelier’s company, Cadens Medical Imaging, with the objective to design Imagia’s EVIDENS platform. In 2019, a consortium led by Imagia and The Terry Fox Research Foundation (TFRI) was awarded up to $49 million by the federal government to establish a Canada-wide health data platform to find personalized cures for diseases.

The award, together with $108 million in cash and $165 million in-kind contributions from 97 consortium partners, will combine Canadian expertise in artificial intelligence and precision medicine to improve healthcare for Canadians and stimulate global commercialization of home-grown research discoveries.

The partners are located across nine provinces and include 31 healthcare institutions, 19 companies, seven universities, 22 research foundations, granting agencies and non-government organizations, as well as all four major Canadian AI research labs.