AI offers hope for more effective depression treatments
Montréal-based Aifred Health is using GPU-accelerated deep learning to predict the best treatments based on a patient’s symptoms, demographics, and the results of certain medical tests.
Thinking about depression treatment is likely to depress you.
The disease, which is the world’s leading cause of disability, affects an estimated 300 million people globally. Although there are plenty of possible treatments, doctors have no reliable way to know what’s best for each patient. Many people struggle with depression for years — enduring medication side effects and feelings of despair — while they contend with what’s now a trial-and-error effort.
“That suffering motivated us to find a better way,” said Robert Fratila, Co-founder and Chief Technology Officer at Montréal-based startup, Aifred Health.
The company is using GPU-accelerated deep learning to predict the best treatments based on a patient’s symptoms, demographics, and the results of certain medical tests. Its work so far has earned it a spot as one of the top 10 teams in the ongoing $5 million IBM Watson AI XPRIZE Competition,
Personalised treatment for depression
Depression is more than just about of the blues. At its worst, it can lead to suicide. When you’re depressed, persistent feelings of sadness or hopelessness can affect many aspects of your life — your interest in everyday activities, how you sleep, your appetite, and even how well you can concentrate.
To treat patients, psychiatrists may choose from dozens of antidepressants, several types of psychotherapy and, in severe cases, brain stimulation techniques. They select treatments based on their experience and medical guidelines, but there’s no objective way to decide, Fratila said.
Aifred Health aims to bring more science into the treatment equation, making it possible for physicians to tailor treatments to individual patients.
“The idea is to give people the right treatments sooner so they get better faster,” Fratila said. That could reduce healthcare costs for treating depression, he added.
Better therapy with biology
For most diseases, doctors can use medical tests like MRIs, X-rays, or blood tests to select treatments and monitor how well patients respond. But no such test is used for depression. That could change as a result of a growing body of research suggesting that neuroimaging, genetics, and other biological factors can point the way to the best therapy.
Aifred Health researchers are combining data from that research with patient demographics, symptoms, and medical history to develop deep learning-based software that would help doctors personalise treatments. They are training their neural network on data from the US National Institute of Mental Health and other sources, using their own cuDNN-accelerated deep learning framework and our GPUs in the IBM Cloud.
In addition, the company is one of 14 teams competing in the SMART Mental Health Prediction Tournament to see who can build the best predictive model for anxiety and depression treatment response.
“With the Tesla GPUs, we’re spending less time waiting for our models to train and more time thinking about how to improve the performance of our networks,” Fratila said.
The company, founded by five McGill University students, is collaborating with medical experts and other data scientists from five universities.
After additional training with more data, the company plans to conduct a series of trials with doctors to test its algorithm against the standard guidelines for prescribing treatments. Trials will explore how well Aifred’s software works, its safety, and whether physicians find it easy to use.
That software will be designed to include checks for drug interactions, health risks and side effects, and also examine how frequently patients are unable to afford recommended treatments.
“This isn’t a replacement for doctors, but a data-powered tool they can use to provide better care,” Fratila said.
(Disclaimer: The views and opinions expressed in this article are those of the author and do not necessarily reflect the views of YourStory.)