Harrison.ai has announced US Food and Drug Administration (FDA) 510(k) clearance for acute infarct triage on its non-contrast computed tomography Brain (CTB) imaging solution.
As per a press release from the company, existing stroke artificial intelligence (AI) focuses on large vessel occlusion (LVO) in one or two vascular territories, with the rest remaining “untriaged”, while Harrison.ai’s triage covers six vascular territories and mechanisms of infarct—prioritising ischaemic strokes “across the full geography of the brain”.
Harrison.ai also says its solution triages based on the “very first scan”—non-contrast head CT—helping ensure suspected ischaemic stroke patients are prioritised for follow-up, including earlier advanced imaging. It triages actual brain tissue injury in anterior cerebral artery (ACA), middle cerebral artery (MCA), posterior cerebral artery (PCA), cerebellar, basilar and watershed infarct locations rather than just vessel occlusions, the company goes on to claim, adding that this US FDA clearance thus represents a “fundamental shift”.
“Acute infarct on non-contrast CT is one of the hardest findings in radiology. Most radiologists will welcome the help in bringing these cases to the front of the queue,” said Jarrel Seah, chief medical officer and practising radiologist at Harrison.ai. “That’s why we felt it was important to build it.”
Harrison.ai states that its acute infarct triage has achieved up to 89.2% sensitivity on thin slices and 85.7% on thick slices, with more than 80% sensitivity and more than 80% specificity on multiple operating points for both thin and thick slice series. Advanced imaging was used as the reference standard for establishing the AI capabilities of CTB, meaning it was validated against confirmed infarcts that “were not guaranteed to be visible” on non-contrast CT.
The US FDA-reviewed 510(k) submission for this clearance included direct performance comparisons against existing stroke AI approaches, the company reports, also noting that the “closest FDA-cleared comparator” on non-contrast CT—an LVO triaging device—demonstrated 63.5% sensitivity and 95.1% specificity for identifying vessel occlusion only. Harrison.ai’s “more comprehensive approach” surpassed 80% sensitivity and specificity over multiple operating points for identifying actual tissue injury across all six vascular territories and mechanisms of infarct, according to the company’s recent release.
The submission also included comparison to a widely deployed, US FDA-cleared CT angiography-based LVO triage system as a reference device. While that system reported 91.3% sensitivity and 85.6% specificity in its most recent clearance, Harrison.ai claims that CT angiography requires contrast injection and is typically obtained after non-contrast CT in the acute stroke workflow, adding that the reference system identifies vessel occlusion but is not indicated for acute infarction triage.
“Patients who aren’t on a dedicated stroke protocol may experience significant delays before acute infarct is identified. When time is brain, that has real patient impact,” said Harrison.ai chief executive officer (CEO) and co-founder Aengus Tran. “Thirteen cleared indications and now our second breakthrough device to reach marketing authorisation; we have the most comprehensive US FDA-cleared AI triage coverage for non-contrast CT Brain.”
