Thomas Insel was formerly the director of the National Institute of Mental Health (NIMH). He works now at Google and recently gave an interview to the New Scientist:
Why did you leave the NIMH to work for Google?
I have to confess that after giving heart and soul to mental-health problems over the last 13 years working in government, I have not seen any improvement for either morbidity or mortality for serious mental illness – so I’m ready to try a different approach.[emphasis added]
Wow. I have never seen a major scientific leader give such a pessimistic summary of his field.
Are you saying Google is a better place to do mental-health research than the NIMH?
I wouldn’t quite put it that way, but I don’t think complicated problems like early detection of psychosis or finding ways to get more people with depression into optimal care are ever going to be solved solely by government or the private sector, or through philanthropy. Five years ago, the NIMH launched a big project to transform diagnosis.
Insel had led an NIMH effort that tried to start over from the beginning on psychiatric diagnosis. He proposed to abandon the purely symptom-based approach of the American Psychiatric Association’s Diagnostic and Statistical Manual. Back when he still had hope for the project, Insel described it this way:
NIMH has launched the Research Domain Criteria (RDoC) project to transform diagnosis by incorporating genetics, imaging, cognitive science, and other levels of information to lay the foundation for a new classification system… This approach began with several assumptions:
- A diagnostic approach [must be] based on the biology as well as the symptoms [of mental illness]…,
- Mental disorders are biological disorders involving brain circuits that implicate specific domains of cognition, emotion, or behavior,…
- Mapping the cognitive, circuit, and genetic aspects of mental disorders will yield new and better targets for treatment.
It became immediately clear that we cannot design a system based on biomarkers or cognitive performance because we lack the data… The diagnostic system has to be based on the emerging research data, not on the current symptom-based categories… We need to begin collecting the genetic, imaging, physiologic, and cognitive data to see how all the data – not just the symptoms – cluster and how these clusters relate to treatment response. [emphasis added]
The problem, however, wasn’t just that NIMH lacked sufficient data to base new psychiatric diagnoses on clusters of abnormal biomarkers and behaviours. Back to the New Scientist interview:
[Insel:] But did we have the analytical firepower to [derive new mental health diagnoses from biological data]? No. If anybody has it, companies like IBM, Apple or Google do – those kinds of high-powered tech engines. [emphasis added]
How can technology help find ways to end mental illness?
It can tell us things that are not obvious from our own eyes and ears. We can now think about using deep learning or intensive data analytics to study behaviour and cognition in a far more objective and precise way. Developing algorithms to decode early changes in speech could help us create devices to identify the early onset of schizophrenia. For example, a group at IBM used speech analytics to identify the first signs of psychosis. The team used machine learning algorithms to identify a particular pattern in the way words were connected. This subtlety hadn’t been picked up by clinicians.
Insel is like St. Thomas More, writing in 1534 from the Tower of London, waiting his execution:
He that in tribulation turneth himself unto worldly vanities,… fareth like a man that in peril of drowning catcheth whatsoever cometh next to hand, and that holdeth he fast, be it never so simple a stick; but then that helpeth him not, for that stick he draweth down under the water with him, and there lie they drowned both together.
(This is the text from which we get our idiom, “to grasp at straws”.) Insel’s vanity is that machine learning will save us by purely empirical discoveries of classifying principles for mental disorders. Here at TIE, we’ve been cautious about the claims for the transformation of medicine via big data. We should try: big data + machine learning will likely help. But in my view, pure empiricism in the absence of a deeper understanding of how the brain works will not be transformative. (Moreover, what’s in it for Google to collect a massive schizophrenia data set?)
We have to persevere, incrementally improving our mental health diagnoses and treatments, while playing the long game to achieve breakthroughs in neuroscience.
