A brighter day: depression diagnosis and treatment

By strengthening and deepening a longstanding collaboration, Norwegian SME, NordicImaging Labs and Amsterdam University Medical Centre in the Netherlands are bringing faster and better diagnosis and treatment for depression.

Depression affects more than 300 million people in the Western world alone. Unlike many other diseases, it is frequently first evidenced at an early age and its debilitating effects are often experienced for the majority of the sufferer’s lifetime. By 2030 it is predicted to be the single biggest cause of loss of quality-adjusted life years. Its societal cost is incalculable.

Innovative collaboration in machine learning

To counter this crippling impact of depression, researchers at NordicImagingLab and Amsterdam University Medical Centre have developed a magnetic resonance imaging (MRI) software programme called DePredict, which accelerates diagnosis and treatment. DePredict uses a machine learning algorithm to predict and monitor the efficacy of antidepressant medication.

At present, between 30% and 50% of first treatments for depression are ineffective and the wind-up and wind-down periods between different courses of treatment can last for three months or more. The efficacy of antidepressant medication has always been difficult to predict. Generally speaking, medication works and performs better than placebo treatments, but we cannot analyse more than this.

DePredict can now predict if depression medication does and doesn’t work. It saves time, money and suffering.

Analysing brain scans

Atle Bjørnerud, of Oslo University Hospital founded NordicImagingLab twenty years ago having already begun working with Professor Liesbeth Reneman of Amsterdam University Medical Centre. The collaboration was encouraged and funded by Eurostars. Now Ole Gunnar Johansen, CEO of NordicImagingLab, and Maarten Poirot at Amsterdam University Medical Centre are at the helm of this research.

At the time, there was little interest in, or funding for, machine based learning research in medical science The computational power needed to process MRI brain scan data was not readily available and there were few available data sets of sufficient size and quality to train algorithms. For context, one MRI scan can generate 5 to 10 gigabytes of data and in a single year, US hospitals generate several petabytes of data. MRI employs magnets rather than the X-rays used in CT which is not harmful and is better at imaging soft tissue and recording its structure, function and blood flow.

At present, DePredict is trained on Sertraline, a commonly prescribed antidepressant drug, but given other suitable data sets, it can also be trained to test the efficacy of other drugs.

“Algorithms are very smart, but they’re also pretty simple; they’re not much use on their own” - Johansen, CEO at NordicImaging Labs

Human knowledge trumps depression

The diagnosis, and even the definition, of depression, is complex and often influenced by many extraneous societal factors, not least, its ambivalent media attention. Johansen is keen to emphasise that while the AI component of DePredict captures headlines, it is human knowledge that is the key to its efficacy in the real world.

When Atle Bjørnerud and Liesbeth Reneman were first beginning their work in the field, some of AI’s most committed evangelists were suggesting that there would be little need for DePredict.

Radiologists interpret images for other doctors and medical professionals. Historically, their work was carried out with the naked eye and corroborated by clinical evidence. Now, by combining multiple images in one algorithm, radiologists have access to a multi dimensional picture and a previously unimaginable quantity and quality of data to share with psychiatrists.

Contrary to the belief that AI would reduce the need for radiologists, as the available supply of data increases and the proliferation of depression continues, there is need for more of them, not fewer.

“If Eurostars hadn’t been set up to encourage collaboration between SMEs and academia across borders then DePredict simply wouldn’t exist.” - Maarten Poirot at Amsterdam University Medical Centre

Interbrain synchrony: when brain frequencies align

As the second generation of collaborators in their respective organisations, Johansen and Poirot demonstrate an easy and engaging familiarity with one another. The former is happy to let the latter, whom he introduces simply as “the scientist”, do most of the talking. Johansen is a keen rock climber and a bungee jumping instructor, and Poirot is an ultra long-distance runner.

Both are eager to talk about their shared interest in neurology and computing science and where and how they interact. They explain the human brain is wired with an interdependent series of networks that govern various functions, such as vision and movement. In patients suffering from depression, a default mode network (linked to human inactivity) is active most or all the time; a likely cause of some of the more common symptoms of depression, such as lethargy and disinterest.

In DePredict, Johansen and Poirot have developed a powerful predictive and diagnostic tool to help psychiatrists treat not only the symptoms of depression but also its causes. DePredict is the product of a longstanding collaboration across disciplines and international borders that reduces both the human misery of depression and its cost to society.

Were you inspired by this story and have a project idea you want to realise?

Through our funding programmes, national/regional funding bodies support SMEs, large companies, universities and research organisations conducting R&D and innovation projects together beyond borders to achieve great results.

Learn more about our programmes and discover whether we have a funding opportunity for your organisation.

Our open calls
Posted 3 June 24