Forward thinkers at High Tech Summit
By Morten Andersen
Forward thinkers at High Tech Summit: Jakob E. Bardram, Professor at DTU Compute and director of the Copenhagen Center for Health Technology (CACHET).

In just a few years, “digital phenotyping” has become a hot topic in health technology. In medicine, the term genotype denotes the expression of a person’s genetic code. The genotype combined with the influence of environmental factors is the phenotype. Ideally, people with the same diagnosis should have different treatment if their phenotypes are different.
“Actually, our behavior is more important than our genetic profile in relation to a number of chronic illnesses. And data which describe our behavior can often be obtained far more easily than genetic profiles, for instance via collecting data from a standard smartphone. This is what we call digital phenotyping.”
Jakob E. Bardram explains.
Speech analysis reveals depression
A joint project with Rigshospitalet supports patients suffering from depression. As mental illness can be associated with lack of self-awareness and loss of initiative, patients may refrain from seeking assistance until the condition has worsened significantly. The idea in the project is for the individuals’ smartphones to sense changes in behavior. For example, the research has shown that a patient may tend to spend more time at home when entering a depressive episode. This will trigger an alarm at the hospital, allowing doctors to approach the person at an early stage, avoiding a worsening of the condition.
“Besides the absence of physical movement, characteristic voice changes can occur. In our research, we were able to detect this while the person makes everyday phone calls. I would like to stress, however, that we are not ease-dropping on the content of the conversation; the system is not listening to the words in the conversation, but merely make a spectral analysis of things like tone and velocity of speech.”
Says Jakob E. Bardram.
Better prognosis with early intervention
Another example is diabetes, where smart phone apps combined with more dedicated wearable devices can supervise modifications in diets and insulin treatment. The result would be semi-automated adjustments, which would optimize the wellbeing of the diabetes patient and minimize side-effects.
While in the cases above, the users have been diagnosed and are receiving medical treatment, Jakob E. Bardram also envisions more general health monitoring through digital phenotyping:
“Relatively subtle changes in our behavior may be early warning signs for a number of medical conditions. For a range of conditions, the prognosis is much better with early intervention rather than late intervention.”
Still, for such applications – as for the depression and diabetes examples – privacy is always an issue.
“We will always need to ensure that projects do not violate the privacy of individuals.”
From phenotyping to treatment
A further question is how to react once digital phenotyping has been completed.
“Right now, our focus is to identify reliable digital markers that allow us to make digital phenotyping. The next step will be to ensure, that the results are actually applied in providing the optimal care considering a person’s phenotype.”
One thing is sure, however:
“The interest in digital phenotyping will continue to grow, and we will of course be discussing the concept extensively at the High Tech Summit 2018.”

“It’s no use closing your eyes to the changes. We need to constantly keep up and try to influence the process.”
Professor Jan Madsen, DTU Compute