The emergence of technologies like ChatGPT has pushed artificial intelligence into the spotlight throughout 2023 - and healthcare is no exception.

The World Health Organization said when launching a set of initiatives: "With the increasing availability of healthcare data and rapid advances in analytical technologies—whether machine learning, logic-based, or statistical—AI tools can transform the health sector."

As we move into 2024, here are some key developments in AI—and warnings—that will be at the forefront of Canadian experts' concerns in the new year and beyond:

Personalized patient care:

Roxana Sultan, Chief Data Officer and Vice President of Health at Toronto Hospital, said one of the most exciting potential developments in AI in healthcare is harnessing the computer model's ability to process and interpret "multimodal" data about the patient.

Sultan also said that currently, AI models can make diagnoses based on one or two pieces of information, such as X-rays. This is achieved by training the model on "tons and tons of X-ray images" until it learns to recognize certain diagnoses.

Sultan said, "This is great. But this is (only) one source of information."

She also added that in the "near future," machine learning will evolve so that AI can take a "more comprehensive look at the patient's health."

In addition to the patient's X-rays, for example, AI will be able to process other data, including doctor's notes, lab samples, medications the patient is taking, and genetic information.

Sultan said this capability will not only play a crucial role in diagnosing the patient but also in coming up with a more personalized treatment plan.

"When you have models understanding the complex interaction between a person's genes and medications and all the different diagnostic tests you perform on that patient, you bring them together into a picture that allows you not only to understand what is happening at this moment but also to plan for the future—if you apply this treatment... what is the most likely outcome for this particular person?"

Agreeing with this view is Ross Greiner, a fellow of the Alberta Institute for Artificial Intelligence.

Greiner, who is also a computer science professor at the University of Alberta, said: "Standard medical practice was one size fits all."

He also added: "Now you realize there are significant differences between individuals... different genes, different metabolites, different lifestyle factors, all impacting (health)."

Machine learning means computers can analyze hundreds or thousands of patient-related features—more than a human doctor can process—and find patterns "that allow us to know that for this specific patient characteristic, you get treatment A, not treatment B," Greiner said.

Clinical trials:

Su Baish, CEO of DIGITAL, one of five "global" Innovation Clusters across the country funded by the federal government, said AI's capacity to handle huge amounts of data will also save "tens of thousands—perhaps hundreds of thousands—of human work hours" for researchers analyzing clinical trial results.

Based in Vancouver, Baish said: "AI can essentially evaluate billions of data points in a fraction of a second."

She continued that this means new drugs can be evaluated much faster for safety and effectiveness.

Improving data quality:

Whether AI is used in clinical care or health research, the results it generates can only be as good as the data fed into it, experts agree.

Greiner said: "Garbage in, garbage out. If you train on bad data, the best I can do is build a model as good as that data, which is a problem."

Sultan said a priority area is ensuring AI receives data from reliable sources, rather than just randomly taking publicly available information. She added that ChatGPT, for example, is a technology "primarily aimed at scraping the internet."

Sultan said, "The problem with that... is first of all that it’s not always reliable or accurate. And second, it’s full of biases and problematic viewpoints that get reinforced when you train something that can’t make those judgments. It just reads everything, absorbs it, and spits it back out to you."

Sultan said one example of how to improve the quality of the medical analyses AI produces is to train it on medical books rather than the internet, adding: "I think ChatGPTs in the world will look like cavemen, as if they are very primitive (in the future)."