Imagia is helping doctors diagnose cancer
For doctors treating cancer, speed is important. The sooner treatment begins, the better the patient’s chances.
But they can’t just rush in. Many tumours are benign, and cancer diagnosis is often invasive and painful.
When a tumour is found in a patient’s lung, for example, doctors will wait until it has doubled in size before ordering a biopsy, said Alexandre Le Bouthillier, the founder and COO of Imagia.
“It can take six months, it can take a year to double in size and then they will investigate further,” he said. That’s a long time to wait if the tumour is malignant.
Le Bouthillier’s company, Imagia, is developing a system to speed up that process. It draws on advancements in artificial intelligence and radiomics, a field of study based around the idea that medical images contain data about the underlying condition, data that can be processed by computers.
“Right now, it’s a race against the clock and AI can really help,” Le Bouthillier said. An AI system can extract more information from an image than the naked eye.
Essentially, Imagia’s AI system looks at pictures or videos of tumours and identifies the microscopic effects of genetic mutations. Using that information, the system is able to predict what’s happening at the genetic level.
Imagia currently has a working prototype of its system attached to an endoscope, a camera-equipped tool used to look inside the human body.
Doctors often use endoscopes to look for polyps, or small growths, during colonoscopies. If they find one, they remove it and send it to a pathology lab for a biopsy.
Imagia’s system can predict whether a polyp is cancerous in real time with more than 90 per cent accuracy.
It’s not a replacement for a biopsy, Le Bouthillier said. Instead, it gives doctors more information, which means they can order a biopsy sooner if a tumour is likely cancerous and, if a tumour is likely benign, they can avoid unnecessary biopsies.
To train the system, which uses a type of artificial intelligence called deep learning, Le Bouthillier and his team showed it images of tumours, along with data about the tumour’s specific genetic mutation, the treatment the patient received, any side effects they experienced and their ultimate outcome.
This was done thousands of times — the deep learning training process involves feeding a massive amount of data into the system and then correcting it through algorithms until it’s accurate.
Because the system is able to deduce what’s happening on the genetic level by looking at tumours, that opens the door for it to be used for personalized medicine. Patients could be given treatments designed for them and the specific type of cancer they have.
“My objective and Imagia’s mission is to make personalized medicine accessible to everyone, because personalized medicine is what is required to fight those diseases that are specific to genetic mutation,” Le Bouthillier said.
Imagia plans to partner with medical device makers and pharmaceutical companies to bring its system to market. Just how quickly it will be used in clinical settings depends on those partnerships and regulatory approval.