Can AI Help To Diagnose Mental Illness?

Last year I wrote about a fascinating project involving researchers from Swansea University.  They were looking to utilize big data to better understand the issues surrounding the impact mental health has on young people.

They analyzed data from 358,000 people aged between 6 and 18 years of age living in Wales between 2003 and 2013.  The data was gleaned from GPs and NHS primary care services.

The data revealed that antidepressant use rose significantly, with depression symptoms doubling in that time.  Interestingly however, actual diagnoses of depression fell by roughly a quarter.

‘These findings add to the growing debate over increasing prescribing of anti-depressants to children and young people. The main issue is whether they being prescribed appropriately. However, it’s worth remembering that there has been historical under- treatment of mental disorders in young people. It’s important that each individual young person is listened to and gets the right kind of help for their problem,” the researchers say.

Utilizing machine learning

Suffice to say, data is at the heart of machine learning, and a recent study from the University of Alberta highlights the promise of automating the diagnosis and treatment of mental health disorders.  The study uses MRI images of people who have been both newly diagnosed with schizophrenia and healthy subjects.  They believe that by examining the connections in a brain region called the superior temporal cortex, they can successfully identify schizophrenia patients with 78% accuracy.  It was also able to predict with 82% accuracy whether that person would respond to specific treatments or not.

“This is the first step, but ultimately we hope to find reliable biomarkers that can predict schizophrenia before the symptoms show up,” the researchers say. “We also want to use machine learning to optimize a patient’s treatment plan. It wouldn’t replace the doctor. In the future, with the help of machine learning, if the doctor can select the best medicine or procedure for a specific patient at the first visit, it would be a good step forward.”

As with so many medical conditions, early diagnosis is key to the successful treatment of schizophrenia, which is estimated to affect roughly 1% of the population during their lives. This allows medical staff to devise a personalized treatment strategy that goes beyond the trial-and-error approach that sadly dominates today.

Many of these data and AI driven projects are at an early stage, and much more work is required before they can make a different to people’s lives.  The Canadian team accept this and aim to further test their approach on a larger sample to try and improve the accuracy they’re capable of achieving, but the initial results are certainly interesting.

“It will be a joint effort of the patients, psychiatrists, neuroscientists, computer scientists and researchers in other disciplines to build better tools for precise mental health,” they conclude. “We have a Computational Psychiatry group at U of A with a team of excellent clinicians and scientists to work collaboratively on this challenging problem.”

Of course, a challenge with their approach is that it requires MRI images to be effective, which may not be available at an early enough stage to be useful.  Perhaps the ultimate solution will combine the approach taken by the Welsh team with that of the Canadian team, so that standard medical record data can provide early warning signs, before an MRI then confirms the suspicions.  It will be interesting to see how both projects progress from here.