The Next Big Thing for Monitoring Parkinson’s Disease
by Jon Palfreman, PhD on 18 April 2016
Neurologists typically monitor their Parkinson’s patients infrequently. Patients like me go in once every 6 months or so and do a test — the Unified Parkinson’s Disease Rating Scale or UPDRS. Every person with Parkinson’s knows this test, which takes about fifteen minutes. It covers fourteen categories where muscles may be behaving badly, including speech, facial expression, tremor, rigidity, finger tapping, hand movements, hand pronation/supination, foot tapping, ability to rise from a chair, posture, gait, and balance. The aggregate score (between 0 and 108) is supposed to quantify a patient’s motor condition and, over time, chart the progression of the disease.
The scale has many critics. While some neurologists and researchers defend it, patients privately mock it as an unrepresentative snapshot of their complex multifaceted disease. As the Swedish patient and advocate Sara Riggare puts it, “I see my neurologist every six months for a thirty-minute session and thus spend every year a total of one hour being observed [with the UPDRS]” That leaves, she says, “8,756 hours per year when my disease is not being monitored.” The test is highly subjective as well. Two neurologists will often give a different score for the same patient on the same subtest. Results vary dramatically depending on how recently patients have taken L-dopa and the time of day. Patients perform up to 30 percent better in the clinic than they do at home. Yet, with all its flaws, researchers are compelled to use it as the controlling metric.
Over the past decade, a number of companies have developed technologies capable of tracking Parkinson’s patients’ symptoms outside the clinic 24/7 — including the Cleveland-based Great Lakes NeuroTechnologies, the Australian company Global Kinetics, Portland-based APDM, and the Cure Parkinson’s Trust–supported European SENSE-PARK project. These entities use various combinations of advanced wearable sensors (worn on sites like the wrist, waist, and ankle), which use accelerometers to track multiple domains round the clock — bradykinesia, tremor, walking, gait, balance, cognition, and more.
While these gadgets are promising, I must confess that I have found them disappointing. The information gathered is hard to interpret and the technology still lacks precision. One device that I tested interpreted a long daily walk as a prolonged episode of dyskinesia. And so far, the devices don’t do a good job distinguishing bradykinesia (slow and impoverished movement) from inactivity (e.g. sitting down reading a book.).
Then just last week IBM and Pfizer announced a new collaboration that takes patient monitoring to a whole new level. IBM’s Watson division—the center with the famous supercomputer that beat out all human competition at Jeopardy — now is directing Watson’s AI skills to health care, with a special emphasis on Parkinson’s disease. Watson can be fed unstructured information — everything from medical research articles to patient records — and learn over time to find patterns and make inferences. This remarkable “machine learning” ability enabled it to master the game of Jeopardy (See Watson’s impressive conquest here). In its quest to revolutionize health care, one of its first targets is Parkinson’s disease.
Pfizer and IBM’s research collaboration to transform PD care will utilize what has become known as “the internet of things;” meaning the hundreds of sensors and mobile devices that are now in the world, everything from thermostats, to light dimmers, to motion detectors. It will use these ubiquitous sensors, along with the wearable sensors mentioned above on people to generate real time information about a patient’s movement, cognition, sleep, along with daily activities like grooming, dressing and eating.
To realize this goal they are building a special instrumented house, decked out with every kind of sensing device at IBMs Watson Research facility in Yorktown Heights NY. IBM’s Ajay Royyuru says they will place a whole array of motion sensors and proximity sensors in this experimental environment. “We can deploy sensors on surfaces that can record an event like a bathroom door being closed or a shower door being closed, or actually even a toilet flush being used. ” Likewise, says Pfizer’s Peter Bergethon, the team plans “to deploy sensors on the furniture and on the floor… those gadgets can tell us whether a chair is being occupied. And we might even be able to tell what is the posture of the person sitting in the chair. So we might actually be able to detect dyskinesia from the sensors on the chair.”
This massive stream of continuous information will be fed into the Watson supercomputer, who will also been stuffed with vast amounts of knowledge about Parkinson’s disease, including clinical records. When actual human subjects — both people with Parkinson’s and healthy controls — spend time in the instrumented house later this year, Watson will be able to monitor their entire life experience on a second to second basis. Watson will look for patterns that will correspond to verifiable clinical end points. As Bergethon puts it, “We want to define the digital signature of a person, then characterize how they are feeling and how they’re responding to medication.”
The plan is to figure out over 2-3 year period just which sensors provide the greatest insight for PD. Once this is known, it should be possible to use these sensors in actual patients homes, providing a vastly superior picture of how a patients is doing, rather than trusting self reported symptoms or a six monthly UPDRS test. If it works it may revolutionize the clinical trial process and the prospects for new treatments.