Language and mental health

How linguistics can predict a patient’s mental breakdown
“Fire. Water. Energy.”
That’s the answer the young woman gave to her mental health counsellor, Purple Weber, when Weber asked, “Why do you think everything is contaminated?”

Weber’s client is being treated for schizophrenia and perceives the world around her as a dangerous and harmful conspiracy. Her delusion is a symptom of her illness − as are the language and syntax she uses.

“She throws different words at me that lack connection,” explains Weber, who works with the Schizophrenia Society of Ontario. “There is cognitive confusion.”

Some other clients, she says, “create words or give new meaning to words they use.” One man with paranoia always uses the words “poker face” to describe the people he believes are following him. “We’re not sure whether it’s the thoughts that are jumping,” explains Weber, “or some of the connections in language.”

According to IBM, each person generates one million gigabytes of health-related data across his or her lifetime. (COURTESY IBM)
Either way, the unusual pattern of communication − “gaps, things missing, incoherence, or connection which is out of ordinary” − is a sign that something in the way the mind works is disorganized.

But often, even in reviewing taped or transcribed sessions, says Weber, “language is a huge piece that we miss.” She adds, “It would be very, very helpful, and a lot more scientific, if there was software to analyze the language used.”

Indeed, there is.

Dr. Guillermo Cecchi, a computational neuroscientist with IBM’s Watson Research Center, has been working on applying computational linguistics, using speech samples, to psychiatric conditions including schizophrenia, mania, and even drug and alcohol intake.

“I interface computer science with a special interest in mental health,” explains Dr. Cecchi.

This has become possible through the augmented intelligence (AI) of IBM Watson, which is able to recognize, understand and communicate using natural language.

Earlier this year, IBM released its Tone Analyzer in beta, which can parse through text to identify the emotions and tone of a message. (COURTESY IBM)
“Speech is a signal,” he says. “Practitioners use it in the field on a daily basis. Now we have tools that are extremely powerful to analyze language, independent of the judgement of one particular physician.”

But Dr. Cecchi and his Watson team are taking computational linguistics beyond diagnosis.

“What I’m most passionate about is the idea of using language as a tool for prognosis,” he says. “When we started to work with some of the first data sets, what we found surprising was that we found the signal very quickly.”

Refining the data and recognizing the complexity, he explains, “we realized from the outset that the signal was predictive.” Simply put, that means Dr. Cecchi’s research showed that Watson can predict a psychotic episode in youths who are at risk and can identify those who are at higher risk.

“You can monitor the patient at a very high frequency,” explains Dr. Cecchi.

He says: “There are remote villages in India, for example, that don’t have any psychiatrists or physicians. This gives the possibility of remote evaluation. A trained psychiatrist in New Delhi would be able to use this and direct a social worker in the village.”

“We see that as the main use of this technology,” he says, “the high frequency, immediate access to the patient, even patients who are immobilized.”

Dr. Cecchi describes it as “the expansion of psychiatry to a much larger population.”

There is, he says, “a dire need for new approaches in mental health. It’s the largest component of health care in most countries. And yet the techniques used for diagnosis are not very different than those that were being used 100 years ago.”

In 2015, Manipal Hospitals in India announced that it would be adopting Watson for Oncology, which will help clinicians analyze massive amounts of data. (COURTESY IBM)
AI vs. arthritis

Augmented intelligence, also known as cognitive computing, is looking to transform many other areas of health care.

At the IBM Life Sciences Discovery Center at Princess Margaret Cancer Centre in Toronto, senior scientist Igor Jurisica has used artificial intelligence and WorldCommunityGrid resources to identify prognostic and predictive signatures for lung cancer and is now applying the technique to ovarian cancer.

The same techniques and data analysis can also be applied to other diseases, he says. “Everything that we developed would be applicable to diabetes, heart disease and arthritis.”

Computational analysis could “bring forward markers to determine who needs what kind of treatment and to triage patients appropriately.”

Dr. Jurisica’s team is currently “fast-tracking” arthritis. The disease, he says, “is costing Canada a lot more than cancer. It’s a long term ailment that people live with for many, many years. Improving diagnosis and treatment could tremendously improve quality of life.”

His lab is using AI to combine, for the first time, three types of information about arthritis patients: Clinical information from patient records; molecular analysis of genes, DNA and microRNA; and data from wearable devices. The goal is to determine which patients with arthritis would benefit most from surgery and those who would benefit equally from physiotherapy.

Dr. Jurisica says adding Watson to the mix will complete the picture by “bringing in all the relevant literature about arthritis.”

Research analyst Dave Schubmehl, who covers cognitive computing at International Data Corporation, says he’s not surprised that IBM is looking well beyond Watson’s pioneering oncology application. “They’ve been working with pharmaceutical companies to help facilitate basic research, to find what substances cause reactions and to determine drug indications, helping them connect the dots,” he says.

ARTIFICIAL INTELLIGENCE
Special to The Globe and Mail Published Thursday, May 19, 2016
source