Photo Source: Google DeepMind
A revolutionary advancement in artificial intelligence has a chance in transforming stroke care. Researchers from Imperial College London, the Technical University of Munich, and the University of Edinburgh have designed AI software that can accurately analyse brain scans to detect the precise moment a stroke occurs.
This significant development in diagnostic assessment, outlined in npj Digital Medicine, promises to reshape how strokes are diagnosed and managed.
The Critical Need for Timing in Stroke Treatment
Strokes are caused by a blockage in blood flow to part of the brain, depriving brain tissue of oxygen and nutrients. Immediate treatment is critical, as delays can make therapies ineffective or even harmful. Clot-busting drugs and surgical interventions are usually effective within the first 4.5 to 6 hours after a stroke occurs.
However, diagnosing the precise onset time is challenging, especially when patients are unable to communicate or when strokes occur during sleep. Traditionally, stroke onset is determined by visually examining CT scans for darkened areas of the brain. This method is subjective and can be unreliable due to variations in brain structure and blood flow.
AI Delivers Precision and Speed
The new AI software was trained on a dataset of 800 brain scans with known stroke onset times and validated on nearly 2,000 additional cases. According to the researchers, the software is twice as accurate as standard visual methods. It evaluates lesion texture and variations, not only estimating the chronological time of stroke onset but also assessing the biological age of the lesions—indicating whether the damage is reversible.
Dr Paul Bentley, who led the study and is a consultant neurologist at Imperial College Healthcare NHS Trust, explained:
“For the majority of strokes caused by a blood clot, if a patient is within 4.5 hours of the stroke happening, he or she is eligible for both medical and surgical treatments. Up to six hours, the patient is also eligible for surgical treatment, but after this time point, deciding whether these treatments might be beneficial becomes tricky, as more cases become irreversible.”
This ability to simultaneously estimate time and assess reversibility offers a significant leap forward in stroke care. Dr Bentley emphasised:
“Not only is our software twice as accurate at time-reading as current best practice, but it can be fully automated once a stroke becomes visible on a scan.”
Opening Up More Treatment Options for Stroke Care
The implications of this technology are profound. Researchers estimate that up to 50% more stroke patients could become eligible for treatment with this enhanced diagnostic capability. This is particularly important in emergency settings, where rapid decisions are critical.
The software accelerates the diagnostic process, helping doctors identify the most appropriate treatments without delay, by automating the analysis of CT scans.
Building Trust in AI Diagnostics
The success of this software will depend on patient and clinician trust, as with other AI applications in healthcare. Transparency and explainability are key factors in fostering this trust.
Unlike opaque “black box” AI systems, explainable AI programmes can provide clear insights into how they reach their conclusions, enabling both clinicians and patients to understand the diagnostic process.
This transparency addresses common concerns about AI reliability. For example, a recent survey of patient attitudes toward AI in radiology revealed that many appreciated the consistency of AI compared to human variability. One participant noted:
“Readings of various scans (MRI, CT, X-ray) are already rather unreliable. Show the same film to 100 radiologists, get at least 10 different answers.”
A Future Shaped by AI
The development of this stroke diagnostic tool aligns with broader trends in AI adoption within radiology and healthcare. In 2023, nearly 80% of AI-enabled medical applications approved by the FDA were related to radiology. AI’s pattern recognition and image analysis strengths make it uniquely suited to fields like stroke diagnostics.
The researchers behind the stroke software hope to integrate it into the National Health Service (NHS) and other healthcare systems worldwide. If successful, this technology could not only improve patient outcomes but also reduce healthcare costs and address critical workforce shortages.