Artificial intelligence could be twice as accurate as current methods in determining when a stroke occurred, which could help doctors make emergency care decisions, according to a study.
Using technology to read brain scans of people who have suffered a stroke could allow up to 50 percent more patients to receive appropriate treatment, research suggests.
The team behind the software claims it is twice as accurate as current methods by which doctors visually assess brain scans to make treatment decisions.
Developers hope that the model, which was created by scientists from Imperial College London, the Technical University of Munich and the University of Edinburgh, will eventually be used in health services.
The algorithm was trained with the help of 800 brain recordings where the time of the stroke was known, and then it was tested on 2000 patients.
The study showed that the model was twice as accurate as the visual method, which the scientists suggested could be due to the fact that it picks up additional features from the footage, such as texture.
Patients who go to the hospital with a suspected stroke are given a CT scan to determine the cause and severity, within an hour if possible.
Doctors then review the results and make decisions about treatment, where time is an important factor.
Thrombolysis, the process of breaking up blood clots with drugs, can be performed within 24 hours of the onset of symptoms.
Thrombectomy – when doctors surgically remove a blood clot to restore blood flow – is most effective six hours after the onset of symptoms, although in rare cases it can be performed up to 24 hours later.
dr. Paul Bentley, from the Department of Brain Sciences at Imperial College, and a neurologist at Imperial College Healthcare NHS Trust, led the study, which is published in the medical journal NPJ Digital Medicine.
He said: “It is crucial for doctors to know the onset time of a stroke, as well as whether the symptoms of a stroke can be reversed. Having this information helps doctors make urgent decisions about what treatments should be undertaken in patients with stroke.”
“Not only is our software twice as accurate at reading time than current best practice, but it can be fully automated once an impact becomes visible on a scan,” it said.
Lead author dr. Adam Marcus added: “We estimate that up to 50 percent more stroke patients could be treated appropriately thanks to our method,” he asserted.