AI in Stroke Detection and Neurology

Ischemic stroke happens when a clot blocks blood flow to part of the brain. The most severe cases involve a large vessel occlusion (LVO), which can be treated by mechanical thrombectomy, a procedure that removes the clot through a catheter. The benefit falls as time passes, which is why stroke teams say 'time is brain.' AI triage software analyzes CT angiography as soon as a scan is complete. In 2018 the FDA granted De Novo authorization to Viz.ai's LVO software, establishing a device category called computer-aided triage and notification (CADt). These tools do not replace the radiologist's read. They run in parallel and send an alert with key images to the stroke team's phones, so specialists can mobilize while the formal report is pending. In 2020, CMS granted Viz.ai's software a new technology add-on payment, the first for an AI tool, signaling that triage AI could be reimbursed. Studies in hospital networks have reported shorter times to treatment decisions, though effects vary by site. Other tools support different decisions. RapidAI's imaging software was used to select patients in the DAWN and DEFUSE 3 trials, which showed that some patients benefit from thrombectomy up to 24 hours after onset when imaging shows salvageable tissue. Brainomix's e-ASPECTS scores early ischemic changes on non-contrast CT. Similar software flags intracranial hemorrhage. In epilepsy, algorithms help clinicians review long EEG recordings, and some wearables are cleared to detect certain convulsive seizures and alert caregivers. Research also applies machine learning to MRI for tracking multiple sclerosis lesions and to movement data in Parkinson's disease. A common misconception is that CADt tools diagnose stroke. They prioritize and notify; if the software misses a case, the scan simply follows the normal reading workflow, and stroke has not been ruled out.