Machine learning offers automated diagnosis in stroke and dementia

Machine learning offers automated diagnosis in stroke and dementia

1200 800 Peter Stevenson, PhD

Researchers have harnessed the power of artificial intelligence (AI) to identify a key cause of dementia and stroke – and with stunning accuracy – a new study reports.

Created by scientists at Imperial College London and the University of Edinburgh, UK, the new technology identifies and measures the severity of small vessel disease (SVD) – a neurological disease characterised by reduced blood flow to the brain, and one of the most common underlying causes of stroke and dementia.

Using machine learning, the software is able to identify SVD markers present in computed tomography (CT) brain scans of patients presenting with stroke or memory impairment, and attribute a severity score to the detected disease.

The software was tested in 1,082 CT scans from 70 UK hospitals between 2000 and 2014, and as the researchers discovered, the process was consistent and highly accurate: when compared to the results gleaned from a panel of expert doctors, the level of agreement of the software was as good as between the doctors themselves. When compared to the current diagnostic gold standard, MRI, the software reached an accuracy of 85% in predicting the severity of SVD.

Crucially, MRI-led diagnostics demand a lot of time from physicians, and it can be difficult to assess the spread of disease with the human eye. In addition, scanner availability, especially in emergency situations, is limited. By utilising CT scanning with added AI capabilities, the technology could be employed to work through huge data sets, identifying and predicting stroke or dementia risk in a wider population of patients, as well as in emergency situations where decision making is paramount.

As such, the software has great potential in personalised medicine, allowing quick and accurate assessment of stroke risk, and giving emergency workers more information on which to base their triage decision. For example, more extensive SVD carries with it an increased risk of haemorrhage, so treatment decisions need to be weighed against the overall risks of more intensive intervention.

In the longer term, the researchers hope that the software can assist in quantifying the likelihood of slow-progressing SVD – a root cause in dementia or immobility – thereby allowing doctors to actively treat any reversible causes such as high blood pressure or diabetes.


Imperial College London. Artificial Intelligence Improves Stroke and Dementia Diagnosis in Brain Scans. NeuroscienceNews. Retrieved May 17, 2018 from


Brainwork is supported by unrestricted grants from: