Artificial Intelligence brings precision diagnosis and personalized treatments to mental health care. 

Scientists at The Royal are leveraging Artificial Intelligence to bring precision diagnosis and personalized treatments to mental health care.  

Using AI to ‘zoom in’ on the brain 

Dr. Georg Northoff, a psychiatrist, neuroscientist and philosopher, is using artificial neural networks in combination with brain imaging to explore and analyse the neuronal activity that defines how our brains interact with the world around us.  

Brain imagining, such as that produced by the PET-fMRI machine at The Royal’s Brain Imaging Centre, enables scientists to see the changes in brain function and activity that occur in the brain when someone is suffering from a mental illness.  

Artificial Neural Networks are complex mathematical and computational tools that capture patterns and find structure with the complex activity of the brain.  

Northoff uses a familiar iceberg metaphor to illustrate how, together, these two tools can enhance our understanding of the brain.  He explains that brain imagining lets us see what is above the water while artificial neural networks enable researchers to go deeper, allowing them to zoom in and explore the relationships between particular aspects of the brain.

Examining the brain in this non-invasive way takes clinicians beyond understanding what changes are happening in the brain when someone experiences mental illness to understanding why and how those changes are taking place.  

“There is a deeper level of the iceberg beneath the water, and that's where the mechanism plays,” says Northoff. “We can feed the brain imaging data of an individual into an artificial network model, and that basically gives us insight into how these changes we see on the brain imaging are caused. We see what are the underlying features, and those features, we can use for diagnosis.”    

This also opens to the door to designing highly effective personalized treatments – not just medications but non-drug treatments like music therapy and breathing techniques – all based on a patient’s own neural activity.

As an example, Northoff recalls a young patient who came to see him with her mother. The patient didn’t speak at all throughout the appointment.  When Dr. Northoff asked her why, she said that she couldn’t follow the speed of the conversation, it seemed to her as though everyone was speaking too quickly.  

This experience of an individual’s thoughts and movements slowing down is a common early symptom of depression.  Understanding what is happening in the brain to cause this slowing down can help make diagnosing and treating depression more efficient.

Northoff explains that brain imaging data would capture the changes that are happening in the brain when these symptoms occur.  He can feed that data into the artificial neural network model to explore cause and effect – the activity and connections between different parts of the brain that result in the symptom occurring.  This knowledge is helpful for both confirming a diagnosis and for tailoring therapy to address the symptoms.  

“Currently in psychiatry, we don't have many objective biomarkers. Diagnosis is subjective based on the physician’s observation. One of the reasons is that we don't yet have knowledge of the underlying mechanisms of various mental illnesses,” explains Northoff. “Diabetes treatment provides a good example of how important that knowledge is. In diabetes, you have all these symptoms all over the body, which are related to glucose, and that in turn is modulated by the pancreas and insulin. Based on this knowledge we know how to treat diabetes by moderating insulin based on an individual’s glucose levels. We are using brain imagining and artificial neural networks to gain similar insights into the brain and tailor treatment for mental disorders.”

In the case of Northoff’s young patient, brain imagining and artificial neural networks could be used to analyze the speed disturbance occurring in her brain and provide insight into how to tailor interventions to help.  The result: AI-guided music therapy that would start at a tempo that her brain could process then gradually be increased to encourage the neuronal activity in the brain to speed up as well. Northoff likens this to physical therapy in which someone who wants to run may start out walking and gradually increase their speed, training their muscles to run. 

Read more about this ground-breaking brain research in this article from the Ottawa Citizen. You can also check out Georg Northoff’s website at

AI-driven tools to identify suicide risk and help people in crisis

Dr. Zachary Kaminsky, a molecular biologist and the DIFD-Mach-Gaensslen Chair in Suicide Prevention Research at The Royal, is employing Artificial Intelligence in his work to identify people who are at heightened risk for suicide.  This creates the opportunity to intervene and provide support that could save lives.  

Dr. Kaminsky has developed the Suicide Artificial Intelligence Prediction Heuristic (SAIPH), an advanced algorithm that analyzes speech patterns in public Twitter data to help predict suicide risk. SAIPH not only assesses an individual’s future risk of suicidal thought, but also when they will be at risk, based on publicly available data in Twitter posts. 

Kaminsky describes AI as a “work horse” for building models that classify data and finding the patterns within it.  That capability is paired with expert knowledge to create highly effective tools for predicting risk. 

For example, Twitter posts can contain all manner of data and insights into users’ behaviour. Instead of simply feeding all social media data into the algorithm, Kaminsky and his team first distilled data into concepts that they thought were important for understanding suicide and used those to build SAIPH. 

Kaminsky notes that the same process can be used to create tools within our health care system.  For example, using artificial intelligence to analyze health record data could provide insights about what treatments will work best for different patients.  

“When it comes to psychiatry, we need to be able to understand how to treat people with what's going to work for them,” says Kaminsky.  He explains that applying artificial intelligence tools to analyse a large data set has the potential to provide that insight into treatment response. In turn, this could be adapted into clinical decision-making tools.

It is important to note, however, that Kaminsky does not anticipate that medical decisions will be made by AI alone. Rather, AI tools will provide insight that doctors can integrate into their individual decision making.

Kaminsky, who has expertise in molecular biology, notes that AI-driven insights can come from a variety of different types of data.  This can be health record data or biological data based on our genetics; and probably, as Kaminsky puts it, “a little bit of both”.  

“We're working on a something in the future that will basically suggest not only who's at risk, but also what to treat them with based on an understanding of different underlying biology,” says Kaminsky. 

Currently many mental health patients experience a period of trial and error before they find a medication that works for them.  AI-based tools, have the potential to help a physician determine more precisely what medication to prescribe or what therapy to recommend.  

“These tools can help find patterns that we haven’t been able to find before so they may lead to streamlining care,” says Kaminsky. “There is hope in AI to do these things.”

Read more about how Kaminsky uses AI to predict suicide risk using social media data in this paper published in npj Digital Medicine (published by Nature).