Imperial-led study delivers fairer remote cognitive testing for stroke

Researchers have developed a way to separate motor delays from thinking time in digital stroke tests, improving accuracy and fairness in assessments.

Researchers have shown that a computer-based approach can separate hand movement delays from thinking ability in self-administered digital tests after stroke, improving validity and widening access to assessment.

A known issue with online assessments for neurological conditions such as stroke or multiple sclerosis is that there is no current way to separate the physical impairments (such as a delay of hand movement) from a cognitive impairment seen in test results. Separating these variables is important, as it helps assess cognitive impairment accurately and observe conditions over time in research and for clinical monitoring.

"We can't change the person's impairments, but maybe we can develop a way to get around those impairments and still get something that is accurate." Dragos-Cristian Guia Department of Brain Sciences

Using a computational modelling framework developed called IDoCT, the researchers analysed each test response and separated the time to think from any delay caused by hand movement or the device.The resulting motor-adjusted scores are fairer and better align with pen-and-paper clinical assessments delivered by a healthcare professional.

“We can't change the person's impairments, but maybe we can develop a way to get around those impairments and still get something that is accurate, and not influenced by motor or physical impairments, when it comes to self-administered digital cognitive assessments” said Dragos-Cristian Gruia, first author and Research Associate, Department of Brain Sciences.

What the study found

The study worked with 171 people who have had a stroke. They were seen soon after the stroke and again later during recovery, and their results were compared with 6,364 healthy older adults. Everyone completed 18 short online tests that looked at everyday thinking skills such as memory, attention, language, problem solving and number skills. The test also checked how quickly and accurately people could point and tap on a screen. These exercises are part of Imperial’s IC3 programme and are run on the Cognitron platform.

In the affected group, participants self-administered the test using the hand that had been affected by the stroke, although not all participants had clinically manifest impaired function. Standard tests showed this motor impairment impacted the test results whereas with the modelling applied to the results, the researchers were able to remove the motor delay from the cognitive results.

This newly derived measure aligned better with activities of daily living, closer to what healthcare professionals observe in practice. People with larger stroke lesions or more white matter changes tended to have lower thinking scores, and this link was still present weeks and months after stroke. This indicates that the measure reflects cognition rather than hand speed or device delays and explained more of the variability in the vascular disease load.

Why this matters, not just in stroke

By adjusting for tremor, slowed hand movements and device lag, the approach makes at-home cognitive tests fairer. It stops mistakenly aligning errors due to impaired hand motor function, to thinking and cognitive impairment, cleanly separates movement from thinking, and lets progress be tracked reliably between appointments. This approach is beneficial in a variety of neurological disorders and is especially relevant in multiple sclerosis, where tremor, muscle stiffness (spasticity) and slower movement often sit alongside cognitive slowing. In other words, patients are assessed on how they think, not how fast they tap. The scores line up better with standard clinic tests, day-to-day functioning, and brain scans, giving clinicians a clearer basis for referrals, treatment optimisation, and trial screening.

What's next

"Our hope is to work with our collaborators to use these digital cognitive tools alongside biological biomarkers to develop a better mechanistic understanding of post-stroke cognitive recovery and vascular dementia." Dr Fatemeh Geranmayeh Department of Brain Sciences

This method is easily applicable to other neurological conditions. Because the algorithm only needs per-item correctness, response time and item identity, it can plug into existing digital platforms without redesigning tests. Taken together, this broadens access now, supports earlier detection, and enables scalable follow-up, which moves suitable check-ins from hospital to home and potentially provides the NHS and researchers with clearer data.

“We took the data from these tests, and we applied our algorithm to separate the cognitive from the motor impairments,” said Dragos-Cristian Gruia. “Our hope is the tool once tested further will improve accessibility to diagnosis and monitoring of stroke patients. While we use stroke as the model in this example, the applications could be wider for other neurodegenerative conditions where we see cognitive deficits such as Alzheimer’s disease and Parkinson’s disease.”

"Our hope is to work with our collaborators to use these digital cognitive tools alongside biological biomarkers to develop a better mechanistic understanding of post-stroke cognitive recovery and vascular dementia " Dr Fatemeh Geranmayeh, Clinician Scientist and honorary consultant neurologist.

Department of Brain Sciences, Faculty of Medicine, Imperial College London; Imperial College Healthcare NHS Trust; King’s College London; Harvard Medical School; University of Edinburgh.

Funded by UK Medical Research Council; infrastructure support from the NIHR Imperial BRC and NIHR Imperial Clinical Research Facility.

Find the publication at eClinical Medicine.