Feature

As the U.S. population ages over the next several decades, the number of Americans diagnosed with Alzheimer’s disease is also projected to rise. One recent study crunched the numbers and predicted that by 2050, 15 million Americans will be living with Alzheimer’s disease, compared with about 6 million today (Alzheimer’s & Dementia, Vol. 14, No. 2, 2018).

Unfortunately, right now, those patients go undiagnosed until they reach the late stages of the disease, at which point it’s too late for successful intervention, says Rhoda Au, PhD, a neuropsychologist in the Alzheimer’s Disease Center at Boston University.

"We’re talking about insidious onset. If you are diagnosed today, you weren’t fine yesterday or a year ago or probably even five years ago," she says.

But those subtle early symptoms are unlikely to bring people into a doctor’s office to be tested, and they might not show up on a standard in-office neuropsychological assessment anyway.

Now, Au and other researchers are exploring how new technologies—including digital pens that record users’ handwriting, mobile activity trackers and smart home sensors—might be able to more quickly catch the cognitive and behavioral changes that indicate the very early stages of dementia in older adults.

Such innovations could allow patients, their physicians and their family members to recognize Alzheimer’s disease and other dementias earlier, giving patients more time to try lifestyle interventions, medication or other treatments. They might also allow doctors to better track how the disease is progressing and whether those treatments are working.

From paper-and-pen to digital drawing

For now, diagnosing Alzheimer’s disease relies—as most disease diagnoses do—on information collected in physicians’ and other health-care providers’ offices. And here, technology can provide a boost to traditional methods.

For example, for decades, one simple cognitive test has been a mainstay in helping to diagnose Alzheimer’s disease and other impairments: the clock-drawing test. In this test, a neuropsychologist or other examiner asks the patient to draw a clock showing a specific time (often 10:11). Healthy patients can do this easily, but those with dementia struggle, drawing clocks with irregular shapes, with uneven number spacing and with hands pointing to the wrong time.

The test is quick and useful, but not perfect. Some research has suggested that it is not sensitive enough to pick up the mild cognitive impairment that precedes full-blown dementia. Also, it relies on the examiners’ subjective judgments of, for example, the amount of distortion of the clock face.

So Au and colleagues at the Boston University School of Medicine tested a digital-age update of the test developed by researchers at the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL). Instead of a simple pen, patients take the test with a digital pen that records their movements and timing as they draw the clock, providing much more information than the finished drawing alone. In a nine-year study with more than 2,600 patients with a variety of diagnoses, including mild cognitive impairment, Alzheimer’s disease and Parkinson’s disease, the researchers found that computer algorithms that used this additional information could provide more accurate diagnoses than the traditional paper-and-pen test (Machine Learning, Vol. 102, No. 3, 2016).

In another line of research, Au is studying how changes in people’s spoken language could someday serve as an early "cognitive biomarker" of dementia. Among her other roles, she is the director of neuropsychology for the Framingham Heart Study (FHS), the long-term population health study that began in 1948 and has now enrolled three generations of participants. In that role, Au has—since 2005—recorded more than 7,000 FHS participants’ spoken responses during standard neuropsychological testing. In one not-yet-published study, she and James Glass, PhD, an expert in machine language learning at MIT’s CSAIL, together with researchers from the company Evidation Health, analyzed years of recordings from 130 patients who were cognitively healthy when they entered the study but went on to develop dementia. They found that changes in those patients’ vocabulary choices, vocal pitch and speaking rate—changes that predated their formal diagnoses—could predict who would develop dementia with more accuracy than looking at their demographic characteristics alone (arXiv, Oct. 20, 2017).

Outside the physician's office

For now, the digital pen and vocal analysis techniques might be used to improve in-office cognitive assessments, but that is just one aspect of what Au and other researchers hope to achieve. More broadly, they’d like to take cognitive assessment out of the doctor’s office altogether, by using smart pens, smartphones, smart-home sensors and other new data-gathering tools to passively detect the early warning signs of dementia as people go about their daily lives.

"When someone has a problem, they usually go to the doctor’s office or clinic, and the doctor takes a history and does some exams," says Jeffrey Kaye, MD, a neurologist and head of the Layton Aging and Alzheimer’s Disease Center and the Oregon Center for Aging & Technology (ORCATECH) at Oregon Health & Science University (OHSU) in Portland. "But that primary-care appointment might take 20 minutes, and a lot of it is dependent on self-report and memory. It’s just a snapshot of what’s really happening. What if instead we could collect information all the time, in the home, in as ambient a way as possible?"

Alzheimer's DiseaseSo, for more than a decade, that’s what Kaye and his colleagues at ORCATECH have been doing. Since 2004, they have installed elaborate monitoring systems in the homes of more than 500 older adults (over age 70) in the Portland area, in a program called the Life Laboratory. All of the adults were cognitively healthy when they joined the Life Lab, but at risk for developing Alzheimer’s disease or other dementia due to their age. The in-home monitoring systems have evolved with improving technology over the course of the study but have generally included motion sensors that can measure things like walking speed, sleep patterns and the time spent in and out of the home; wireless scales to monitor body composition; medication dispensers that report medication-taking activity; software to track participants’ computer use; and more.

After installing the systems, "we say to [the participants], go ahead and live your lives. And then from time to time, we ask them, with their permission, to use the monitoring systems to participate in new studies," Kaye explains.

Over the years, Kaye and his colleagues have used data gathered from these sensor systems to search for behavior patterns that reflect cognitive decline. They’ve found evidence that the speed with which participants move through their homes (Neurology, Vol. 78, No. 24, 2012), the timing of when they take their medications (Alzheimers & Dementia: Diagnosis, Assessment & Disease Monitoring, Vol. 6, 2017), their driving behaviors (Journal of Alzheimer’s Disease, Vol. 59, No. 4, 2017), their sleep patterns (Alzheimer Disease and Associated Disorders, Vol. 28, No. 2, 2014) and other factors are associated with cognitive impairment.

These sensor systems can provide an early warning of Alzheimer’s disease, and they could also give researchers and physicians ongoing feedback that would help them determine whether treatments are working or if patients’ care needs to be increased.

Meanwhile, two of Kaye’s former colleagues at OHSU, Misha Pavel, PhD, an experimental psychologist and electrical engineer, and Holly Jimison, PhD, whose background is in medical informatics, founded the Health At-home sensor systems give researchers and physicians ongoing feedback on patients to help them determine whether treatments are working or if patients’ care needs to be increased.

Tracking Alzheimer’s Disease

Behavior Informatics Lab at Northeastern University in Boston in 2014. They are setting up a similar network of in-home monitoring systems around Boston and are focusing on recruiting a sample of lower-income and diverse older adults.

Pavel and Jimison are also interested in how computer, tablet and smartphone use could provide insight into their participants’ cognitive abilities. They’ve developed a suite of cognitive computer games inspired by neurocognitive tests that they install on participants’ computers, and then allow the participants to play as often or infrequently as they wish. In one study, Pavel and Jimison found that they were able to predict participants’ performance on the standard pen-and-paper Trail Making Test by modeling participants’ interactions with one of their cognitive computer games (IEEE Journal of Biomedical and Health Infomatics, 2014). The game suite also includes games that assess memory, verbal fluency and attention.

The advantage of these home-based computer games, Pavel says, is twofold. First—like Au’s digital pen—they provide information about participants’ mental processes as they play the games, not just how well users did playing the games. Take, for example, something like the online card game FreeCell, which Pavel and Jimison have also used in their work, and in which participants must make a series of logical decisions.

"We can go inside [FreeCell], and at every move we compare the performance of the individual to the move an ideal processor would make," Pavel says. "We can look at whether the moves within the game were rational and as good as possible."

Second, the games provide a window into participants’ cognitive function over time, not just the snapshot of a one-time, in-office assessment. "For the first time in human history we have the opportunity to take objective measures of people’s behaviors in real life," Pavel says.

Help or intrusion?

Like most opportunities, however, these types of monitoring systems will come with costs, and one of the most obvious costs is privacy—a factor the researchers are well aware of.

"In every single study, we debrief people on privacy," Jimison says. They discuss with participants the types of data that will be collected and who will have access to it. In some of their studies, for example, the researchers send summary data to participants’ family members, but only with their permission. "Some people might want that information to go to one adult child, for example, but not the other," Jimison says.

The informed consent conversation doesn’t end after the study begins. "We check in after a year because privacy preferences may change as people begin to understand the system better."

Broadly speaking, Jimison and the other researchers say, many older adults are willing to make a trade-off—giving up some privacy so that they can age in place in their own homes and help maintain their cognitive abilities as long as possible.

Au, too, is now working to bring her research out of the lab and into people’s daily lives. In a study that began last year and will continue through at least 2020, she has recruited more than 2,200 Framingham Heart Study participants to use wearable devices to constantly monitor sleep, heart rate, balance and more to look for early signs of cognitive impairment.

"Eventually, maybe, we can become agnostic to ‘tests’ altogether, and find the set of metrics that are common to all tests, and reflect some sort of cognitive signal," she says. "It will take a while, so my strategy is, in the clinic I’ll do what everybody does, but at the same time try to collect information digitally. And then I’ll also look for ways to pick things up with wearables, so I can start to anchor things that I see in the natural environment. And that’s how we’ll push it forward." 

Further reading

How Technology Is Reshaping Cognitive Assessment: Lessons From the Framingham Heart Study
Au, R., et al. Neuropsychology, 2017

Pervasive Computing Technologies to Continuously Assess Alzheimer’s Disease Progression and Intervention Efficacy
Lyons, B.E., et al. Frontiers in Aging Neuroscience, 2015

Using Behavior Measurement to Estimate Cognitive Function Based on Computational Models
"Cognitive Informatics in Health and Biomedicine," Patel, V., et al. (Eds.) Springer, 2017