AI’s Promise in Healthcare: Transforming Towards More Effective and Efficient Medical Solutions

Artificial intelligence (AI) is already present in our healthcare system, to varying degrees. The problem is that it’s mostly being used by individual care and research units operating independently of each other. According to entrepreneur and Université de Montréal alumnus Alexandre Le Bouthillier, AI solutions need to be widely implemented to increase the speed and efficiency of care.

Le Bouthillier has always been an innovator. Throughout his studies, from undergraduate to Ph.D. in UdeM’s Department of Computer Science and Operations Research (DIRO), he also ran Planora, a company that combined AI with work optimization. He cofounded it with Louis-Martin Rousseau and they were subsequently joined by Jean-François Gagné, the founder of Element AI.

In 2012, at the age of 37, Le Bouthillier sold Planora and took a three-year sabbatical. He was at a crossroads. His father had received a terminal cancer diagnosis, which spurred Le Bouthillier to explore how AI could make the healthcare system more efficient.

In 2015, under the guidance of his DIRO colleague Yoshua Bengio, who was doing groundbreaking work in AI, Le Bouthillier teamed up with Nicolas Chapados, then a Ph.D. student, to found Imagia, a company that offers precision solutions in oncology.

“Reading about oncology, I realized that between the patient examination and the first treatment, there are a number of human interventions that slow down the process,” Le Bouthillier explained. “AI can speed things up by analyzing images of different types of cancer and associating them with specific genetic profiles to determine the best treatment for each patient.”

To date, AI breakthroughs in oncology have mostly involved cancers such as breast, colorectal and lung for which there is a screening system in place and hence large quantities of data available. Despite increasingly effective screening programs, however, the treatments only work in about 30 per cent of cases.

A socially responsible investment

As cofounder and partner in Linearis, a socially responsible fund that invests in AI health solutions, Le Bouthillier is now working to increase the presence of AI in the field.

For example, metabolomics—the science that analyzes metabolites, including cholesterol—will increasingly be used and coupled with other omics, such as genomics.

Using low-cost collection devices, people will be able to provide samples themselves at the pharmacy. The data will then be transmitted to a laboratory equipped with an AI system that within seconds analyzes the person’s condition or response to treatment and recommends any necessary adjustments.

Le Bouthillier cautioned, however, that the healthcare network will need to move beyond fax machines – still widely used in Quebec and elsewhere – for AI to be more widely practicable.

“AI could be used to improve healthcare and make it more accessible by enabling people to do an initial noninvasive screening themselves, close to home and at low cost,” he said. “And on research, replacing mice with human cell lines and organoid models will accelerate the discovery of new treatments.”

Mental health to benefit, too

Psychiatrists and mental-health professionals could use artificial intelligence to personalize care and therapy, as well. A year ago, an article in Proceedings of the National Academy of Sciences reported on a new computational model of the human brain that could improve our understanding of how complex cognitive abilities, including consciousness, develop.

The model was developed by Guillaume Dumas, a professor in Université de Montréal’s Department of Psychiatry and Addiction and head of the precision psychiatry and social physiology lab at the UdeM-affiliated CHU Sainte-Justine Research Centre. Dumas, who has a transdisciplinary background in fundamental physics, systems engineering and cognitive science, has spent over a decade researching the social dimension of human cognition.

In 2010, Dumas used a hyperscanner, a device he invented that simultaneously records the brain activity of multiple subjects, to demonstrate that during social interactions, brains literally get on the same wavelength. They synchronize, in a way.

Building on that discovery, in 2012 he ran the first neurocomputational simulations of two interacting brains, showing that anatomical similarity partly explains these intercerebral synchronizations. Conversely, anatomical dissimilarities could lead to difficulties in synchronizing with others.

In 2014, Dumas drew on dynamical systems theory to introduce a new paradigm of human-machine interaction, in which the human user interacts with a bio-inspired avatar in real time.

“Using this paradigm, I discovered how our neural networks, which register information from our own behavior and that of others, also create connections between the sensorimotor and representational dimensions of social cognition during real-time interactions,” Dumas explained.

By combining computational psychiatry, precision medicine and neuro-inspired artificial intelligence (neuro-AI), Dumas is working to develop a more personalized and predictive approach to psychiatry and mental health.

“In psychiatry, for example, treatments are decided on a case-by-case basis by the attending physician,” he said. “With precision psychiatry, we will be able to rely on objective markers to guide clinical decision-making by psychiatrists and other mental health professionals.”

A ‘digital twin’

Dumas’s computational model of the human brain is part of an effort to improve the use of data in clinical decision-making by creating a “digital twin” of each patient’s brain.

It’s also part of a wider scientific trend: the convergence of neuroscience and artificial intelligence. In fact, his model exemplifies the kinds of biological mechanisms and cognitive architectures that could drive the next generation of artificial intelligence and perhaps even lead to “artificial consciousness.”

“To reach that stage, we’d need to incorporate the social dimension of human cognition, which is what we’re aiming to do next,” said Dumas.

“We’ve already conducted the first biophysical simulation of two interacting brains, which has deepened our understanding of the fundamental mechanisms underlying human cognition and cognitive disorders such as autism.”