Can your typing patterns reveal if you have Alzheimer’s?
The neuroQWERTY technology can analyse people’s typing for signs of motor conditions like Parkinson’s disease – and now the team behind it is looking to expand into the notoriously-difficult area of dementia. Teresa Arroyo-Gallego tells us about the science behind the technology and how the team hopes to reshape cognitive screening.
While many disease areas are increasingly embracing digital forms of diagnosis and treatment, neurology is still largely stuck in an “analogue era”, according to nQ Medical’s chief data scientist Teresa Arroyo-Gallego.
“Neurology is particularly slow in digital adoption, partly because it is not as well understood as diseases like oncology and diabetes, and there are fewer standards against which to validate technologies,” she says.
Today, Arroyo-Gallego notes, cognition is mostly assessed through subjective examinations and patient-reported outcomes.
“There are questions like, ‘Can you name these pictures from different animals?’ or, ‘Can you draw a clock?’. Depending on the way the clock is drawn, there are different clinical interpretations that are derived.
“These tests are not only very subjective, but also expensive and time-intensive since they require an expert to be with the patient. This also means they can only be conducted in a clinic, and the patient will be aware they’re being monitored, which can introduce noise to the results.”
As a researcher in the biomedical engineering lab of the Massachusetts Institute of Technology, Arroyo-Gallego was part of a team who developed the neuroQWERTY technology to detect early signs of motor decline in diseases like Parkinson’s by monitoring how people type on their phones or computers.
In Parkinson’s specifically, the technology focuses on analysing the time between pressing and releasing a key – since this kind of metric is more likely to be controlled by subcortical processes and cannot be ‘faked’ for the seven to ten hours a day most of us interact with our personal devices – but the team says neuroQWERTY can measure a wide range of metrics beyond this.
Now spun out into the company nQ Medical, Arroyo-Gallego and the team are hoping to translate their success in Parkinson’s into one of the most notoriously-difficult disease areas in life sciences – dementia and Alzheimer’s Disease – and bring cognition screening into the digital era.
Signs & symptoms
As Arroyo-Gallego notes, cognitive decline was always going to be a harder metric for nQ to analyse than motor decline, as it is not as well defined.
Nevertheless, nQ knew that typing is governed by both motor and cognitive processes.
“The initial hypothesis was very similar to the one in Parkinson’s – i.e. if there’s any damage or any decline that affects those processes, that should be in some way reflected in the way we type,” she says.
The researchers started with a discovery study to assess the tech’s ability to identify a specific typing signature for mild cognitive impact (MCI).
“We asked the patients to go through a series of semi-controlled typing tasks, both on mechanical keyboards and touch screen devices that were designed to mimic natural use of the device.”
Ultimately the team, working with Dr Luca Giancardo at UTHealth, decided to take an approach that was similar to Parkinson’s disease but with the addition of five extra typing metrics.
These include analysing the pauses between typing different language events/units, the semantic and syntactic complexity of what is being typed, and the keystroke dynamics based on the physical position of the key on the keyboard.
“Even when we had a very small subset of data we were able to achieve a very good separation between what we defined as cognitively normal and cognitively impaired,” says Arroyo-Gallego, discussing the preliminary results of a recent trial for the technology.
“The performance was comparable to the Mini-Mental State Examination (MMSE), one of the standards for cognitive screening, and similar to the Montreal Cognitive Assessment (MoCA), which is the gold standard for screening in Alzheimer’s disease.
“We’ve seen that we’re not only able to detect cognitive decline but we are able to break down those typing patterns to specifically assess how different aspects are affected by the disease.”
From that, nQ believes they can derive biomarkers that are specifically optimised to assess aspects like attention, verbal memory, or nonverbal memory.
“We want to close the loop from typing to the brain and provide a better understanding of cognitive function,” says Arroyo-Gallego.
If the neuroQWERTY technology can prove itself in dementia, then Arroyo-Gallego hopes it will signal a shift away from subjective, analogue tests and provide a more objective, granular, sensitive, and passive way of monitoring all aspects of cognition.
“We want to give physicians visibility on what is really happening to the patient and how they are evolving, rather than just a snapshot from a single clinic visit,” she says.
And this has implications beyond screening and into the elusive area of dementia treatment – as Arroyo-Gallego says that one of the main challenges of finding new Alzheimer’s treatments is that current standards of assessment have difficulties showing a significant improvement over short periods of time.
“There might even be existing Alzheimer’s treatments out there that work, and we just haven’t been able to measure their positive impact properly,” she says.
The technology also opens up the possibility of remote, self-monitoring for patients – which has become increasingly important in the wake of the COVID-19 pandemic, as acceptance of digital technology increases, and physician’s workloads become overwhelming.
“Some patients have mentioned that they would like to see how their daily activities contribute to their brain health and having an objective metric may help them build better habits,” says Arroyo-Gallego.
The team is also looking into whether the technology can help with screening in other disease areas, including ALS, multiple sclerosis, concussion or cancer-related cognitive impairment.
“Enabling precision medicine and personalised care is something that is critical in many of these conditions diseases,” she says, “as they vary massively on a case by case basis.”
About the interviewee
Teresa Arroyo-Gallego is chief data scientist at nQ Medical and a machine learning and signal processing researcher. Her work focuses on the development and application of artificial intelligence methods and systems to solve problems in the biomedical field. In 2019, she was included in the MIT Technology Review’s Innovators Under 35 list.
About the author
George Underwood is the editor for pharmaphorum’s Deep Dive digital magazine. He has been reporting on the pharma industry since 2014 and has worked at a number of leading publications in the UK. He can be contacted on LinkedIn.