Why is this research needed?
Stroke strikes every 5 minutes in the UK and it’s crucial that these stroke patients are treated quickly and effectively to reduce the likelihood of death and severe disability. However, it can be a challenge to quickly identify that someone is having a stroke and the underlying cause. This is a crucial first step in getting them effective emergency treatment and it relies on interpretation of brain and blood vessel scans.
However, in very busy hospitals, there may be delays and scans can sometimes be hard to interpret even for the most experienced professionals. Artificial intelligence (AI) computer software can assist humans to interpret brain scans and they are increasingly available for use in healthcare. More research that is also independent of the companies producing the software is needed to ensure that they do increase the accuracy and speed of diagnosis, they are safe and good value for money.
What will the researchers do?
The researchers will test an AI software, called e-ASPECTS, developed by Brainomix Ltd Oxford UK to understand if it should be used for diagnosis of stroke in hospitals.
1. Test the accuracy of the e-ASPECTS software
The team will look at how well the software can identify and quantifying damage to the brain caused by stroke compared to a human expert, as well as how well it identifies underlying causes of stroke symptoms e.g. stroke caused by blood clot vs a bleed in the brain.
They will also look at if certain clinical (eg. patient age, stroke severity) or imaging characteristics (eg. stroke type, position of patient in the scanner) have an effect on differences between the software and human diagnoses.
2. Understand benefits of using e-ASPECTS software in hospitals
The team will determine if using the software can save time in diagnosis of stroke compared to various types of clinicians involved in stroke care. They will also understand if using the software increases confidence of clinicians and influences their decisions in diagnosing stroke.
How will the researchers achieve their aims?
The researchers will first test the accuracy of e-ASPECTS software by collating and processing thousands of CT brain scans using the software. The AI results will be matched with clinical and radiological data and expert interpretation of imaging to calculate the accuracy of results. As the study uses existing data, it is very cost-effective compared to recruiting new patients to gather data.
The researchers will also work with Brainomix, the company that developed e-ASPECTS to make a test site for assessing how use of the software may alter clinical practice. Clinical professionals from many different countries will be shown CT brain scans with and without e-ASPECTS interpretation guidance, and be asked to make treatment decisions in situations that replicate what they may experience in hospitals.
What did the research achieve?
The researchers adapted their plans to get the best results. In response to software updates, they included testing of e-CTA which adds information from CT scanning of blood vessels and testing of e-ASPECTS’ ability to detect haemorrhage as well as ischaemia. They also included many more brain scans than originally planned.
The researchers collated over 4000 patient brain scans from 9 studies, making this the biggest ever analysis of this software. This included 4100 non-enhanced CT brain scans for testing e-ASPECTS and 668 blood vessel scans for testing e-CTA. The final results for software accuracy will be published soon. Testing with professionals is ongoing.
What is the benefit of this research?
This research sets a precedent for independent testing of AI software in stroke, which can improve stroke care by ensuring the best software is used. The results of this research and similar studies will help NHS trusts and other organisations to make informed decisions when purchasing AI software for their patients.
Dr Grant Mair, lead research on the study said: “By funding this study, RITeS, the Stroke Association have shown a commitment to improving front-line care for people presenting to hospital with symptoms of stroke. This funding shows a willingness for the Stroke Association to engage with cutting-edge and potentially transformative AI technology being developed with the aim of making emergency stroke care more efficient.”