RADAR-AD are experimenting with different devices and software to track functional and cognitive decline in Alzheimer’s
The team behind the RADAR-AD project is building a digital platform that will bring together smartphones, wearables and smart home devices to track changes in the behaviour and movements of people with Alzheimer’s. The project results will ultimately help produce more sensitive and user-friendly ways to track disease progression and figure out whether treatments are working.
As part of this mission, the project partners have been testing combinations of technologies that feed digital biomarker metrics to a software application to help analyse a person’s level of brain functioning, basing their research in three different real-life settings: a smart-equipped home, a regular home, and a typical lab visit. Some of the researchers involved in the studies told IMI about their experiences using the equipment and their plans for next steps.
Using sensors to measure gait
For people living with Alzheimer’s, gait has a significant part to play in their ability to continue to live independently. There is currently very little insight, however, into gait impairments during the early, pre-symptomatic stages of the disease that can indicate that there is a problem.
One of the devices tested by RADAR-AD researchers is called Physilog (Gait Up), a sensor-based device that collects metrics on a person’s walking patterns and speed. The researchers are comparing the gait of healthy controls, people with pathological hallmarks, and people with confirmed cognitive impairment.
All participants are asked to perform different walking tasks while wearing Physilog sensors on their feet and one hip, which measure gait speed, steps per minute, and time spend on both feet while walking, which is a measure of stability, along with other metrics. The participants "often wonder why they have to do a walk test, they say ‘what has dementia to do with gait?’, but they understand after explanations,” according to Marijn Muurling, PhD student at the Alzheimer Center Amsterdam (Amsterdam UMC).
“From our experience, the dual task is difficult for participants with dementia. Although they usually can count backwards from 100 to 0 without any problems, and walk fine during the normal walk test, counting backwards from 100 to 0 while walking is difficult for them.”
According to Jelena Curcic, Senior Data Scientist at Novartis, “From the analytics perspective, the IMU sensors work as expected. We did experience some missing data due to app or sensor update or incomplete tests. Also, in the beginning a sensor fell off the foot due to difficult placing of the sensors on the patients due to social distancing and COVID restrictions. But such things happen and they are expected for a small number or participants.”
“Also during analysis we identified that the path of walking was different between sites. Once that was noticed, we explained the importance of the path being the same at all sites and I think that was implemented thereafter. All in all, we believe dual tasking and timed mobility tests are very valid functional tests that will show good results once we have larger sample size.”
The researchers are bringing on board more people to expand the dataset they already have so that they can look for causal links between disease and altered walking patterns. Their findings will also be matched with subjective measures of mobility as reported by the patients themselves.
Using virtual reality to measure planning and navigation
Altoida is an augmented reality app that combines the virtual and real worlds to assess a person’s spatial navigation, memory, planning skills, and their ability to use technology. The app features task that allows it to collect data on fine motor skills, and two hide-and-seek tasks. Hide-and-seek tasks involve requiring people to hide virtual objects in a room and then looking them, a simulation of a typical but potentially complex daily activity.
During these tasks, the Altoida app assesses the movements of the participant as well as how, when and where they place and find objects. “Participants are used to paper and pencil tests for assessing their cognition, so they have to get used to this new kind of test on a tablet. After they are used to it, the majority of participants likes it,” according to Marijn Muurling.
“The software requires a room with a lot of contrast (for example the edge of a chair or magazines on a table), which is sometimes difficult in a ‘clean’ hospital room. Participants with cognitive impairment sometimes don’t ‘get’ the task of placing virtual objects in the real-world environment. For example, if the app asks to place the (virtual) phone in the room, the participant searches for his/her own phone.”
“It also depends on the technology experience of the participant.”
Using banking tasks to measure cognitive function
The ability to manage one’s finances is another vital aspect of daily living, and it’s usually measured via traditional pen-paper methods and interviews. RADAR-AD are testing a banking app that can simulate the experience of using an ATM machine: the user enters a PIN, an amount to withdraw and confirms all inputs using a number pad on a tablet’s touchscreen that resembles an ATM.
The metrics collected include the time it takes the participant to complete the task, and whether or not they got the steps right i.e. did they enter the correct PIN number and amount, and did they hit ‘confirm’?
According to Thanos Stavropoulos, Postdoctoral Research Associate at the Information Technologies Institute (ITI) of the Centre for Research & Technology Hellas (CERTH), “Regarding the banking app, we noticed that the participants performed in a way that was ‘too controlled’ than in previous experiments (just one attempt and no experimentation or delays). This still yields results, but in future experiments we would like to encourage more free experimentation by the participants.”
Marijn Muurling added that almost all participants joke that it would be great if there would come real money out of the tablet after doing the test. “The test is very short and they are usually surprised that it’s already over. Since the app is built on an internet web page, it depends on the internet connection how well the ‘buttons of the ATM’ react to the participant, which might influence results.”
Early data from these experiments will be investigated in more depth as the study and data collection progress.
Linking digital measures with standard testing outcomes
The researchers also performed modelling based on data from the Altoida virtual reality app and data from ADNI, an initiative in which researchers collect data from MRI and PET images, genetics, cognitive tests, CSF and blood biomarkers to make predictions about the disease.
They were able to reveal associations between different types of digital measures across different neurocognitive domains that were statistically significant and predictable, and their analysis revealed links between digital measures and Mini–Mental State Examination (MMSE) scores (the MMSE test is a 30-point questionnaire used extensively in clinical and research settings to measure cognitive impairment) as well as indirect links to imaging and cerebrospinal fluid biomarkers.
The results are preliminary but promising. The researchers will continue to enrol more participants in the coming months and continue their analyses.