Nonlacunar covert brain infarcts on noncontrast CT as sentinel imaging markers of stroke etiology and outcome

Nonlacunar covert brain infarcts (CBIs) on noncontrast CT (NCCT) are frequently overlooked yet may signal high-risk etiologies such as large-artery atherosclerosis (LAA) or cardioembolism (CE). We investigated the clinical-etiological significance of nonlacunar CBIs and developed a deep learning model for their automated detection. Among 628 patients with first-ever symptomatic ischemic stroke from a multicenter registry, nonlacunar CBIs were identified in 13.5% and were associated with atrial fibrillation (adjusted OR 2.03; p=0.002) and with stepwise worsening large-artery disease, manifested as either severe stenosis (adjusted OR 2.48, p=0.015) or arterial occlusion (adjusted OR 2.99, p=0.004). Etiological concordance was substantial and statistically significant. Among CBIs with relevant stenosis, LAA was the predominant subsequent etiology (48.5%, p=0.019), and among those without relevant stenosis, cardioembolism predominated (46.2%, p=0.011). Patients with nonlacunar CBIs had worse 3-month functional independence (51.8% vs 73.1%; adjusted p=0.002) and impaired recovery trajectories (p<0.001). A deep learning model, developed on a separate cohort (n=758) and validated across three independent external cohorts (n=1,680), achieved sensitivities of 0.722–0.755 and specificities of 0.797–0.932. These findings reframe nonlacunar CBIs as sentinel markers of persistent high-risk pathology, and automated detection on routine NCCT may help identify patients who warrant etiologic evaluation and targeted secondary prevention before a subsequent disabling stroke of the same underlying etiology occurs.