The role of computed tomography in thrombus characterization: Insights from a stroke cohort of known and unknown etiology

Introduction: This study aimed to investigate: (1) the correlation between stroke cause and thrombus computed tomography (CT) characteristics in acute ischemic stroke (AIS) patients and large vessel occlusion (LVO); (2) whether thrombus features in cryptogenic stroke (CS) can help differentiate cardioembolic from non-cardioembolic causes. Patients and methods: In consecutive AIS patients with LVO recruited retrospectively (2022–2025), associations between stroke etiology and thrombus CT characteristics (location, density and length) were estimated. The association of cause with thrombus location was estimated with multivariable logistic regression and presented as odds ratio (OR). The association of cause with Hounsfield Unit (HU) and the thrombus length was estimated with multivariable linear regression and was presented as coefficients (ß). A latent class analysis was performed in CSs to identify clinically meaningful sub-phenotypes based on demographic and thrombus characteristics. Results: Among 580 patients (mean age, 75.8 ± 12.4 years; 274 women), 26.2% had large artery atherosclerosis, 58.8% cardioembolism (79.2% of these with atrial fibrillation), 5.7% other determined causes, and 9.3% CS. Non-cardioembolic strokes were associated with a more proximal thrombus location (OR 2.00 [95% CI 1.15–3.44]), higher HU (ß 2.63 [95% CI 0.20–5.06]) and longer thrombus (ß 1.23 [95% CI 0.66–1.79]) compared to cardioembolic strokes. Thrombus characteristics for CS were similar to those of cardioembolic strokes. Within the group of CS patients, two sub-phenotypes emerged: one showing higher thrombus density and length, and another mirroring cardioembolic characteristics. Conclusion: Within the cohort of patients with CS, thrombus characteristics allowed to identify patients with a stroke likely of cardioembolic origin.