Wang, Shunli

Background and Purpose— Nearly 30% of large vessel occlusion acute ischemic stroke clots are from an unknown source. We assessed histological clot composition in a series of patients with large vessel occlusion and investigated correlations between clot composition and stroke pathogenesis. Methods— As part of the multi-institutional STRIP registry (Stroke Thromboembolism Registry of Imaging and Pathology), consecutive emboli retrieved during mechanical thrombectomy were stained using Martius Scarlett Blue and analyzed using machine learning software. We assessed proportions of red blood cells, fibrin, platelets, and white blood cells. Correlations between clot components and stroke pathogenesis (large artery atherosclerosis, cardioembolism, and stroke of undetermined pathogenesis) were assessed using SPSS22. Results— One hundred five patients were included. The proportion of platelet-rich clots (55.0% versus 21.2%; P =0.005) and percentage of platelet content (22.1±4.2% versus 13.9±14.2%; P =0.03) was significantly higher in the large artery atherosclerosis group compared with the cardioembolic group. The proportion of platelet-rich clots (50.0% versus 21.2%; P =0.024) was also significantly higher in the cryptogenic group compared with cardioembolic cases. Large artery atherosclerosis and cryptogenic cases had a similar proportion of platelet-rich clots (55.0% versus 50.0%; P =0.636). There was no significant difference between stroke pathogenesis and the other major clot components. Conclusions— High platelet content of emboli is associated with a large artery atherosclerosis etiology of large vessel occlusion.

Introduction: The composition of acute ischemic stroke (AIS) blood clots can impact their successful removal via endovascular therapy. Identification of the clot composition prior to endovascular treatment may influence the optimal treatment strategy. In this study, we aimed to investigate the relationship between the Hounsfield Unit intensity on CT and machine-learned characterization of clot composition. Methods: Prior to endovascular treatment, each patient had a non-contrast CT performed. Histopathological analysis of the clots was performed and clot composition was quantified via standard thresholding techniques using Adobe Photoshop. In addition, following representative cell labeling by an expert pathologist, image segmentation algorithms (Orbital Image Analysis, Actelion Ltd.) were trained on each case to quantify composition by cell type. The clots were categorized into 3 groups: red blood cell (RBC) dominant (>50%), RBC proportion equal to fibrin and fibrin dominant (>50%). Correlations between clot composition and Hounsfield Units (HU) density on CT were assessed. A positive Hyperdense Artery Sign was defined as ≥50 HU. Results: Twenty patients were included in the study. There was a significant correlation between clots quantified using Adobe Photoshop Image-Pro and Orbit image analysis ( ρ = 0.837, p < 0.001). A significant relationship was found between clot composition and the presence of a HAS sign, ( X 2 (2) = 6.496, p = 0.039). There is a significant positive correlation between percentage of RBC in the clot and the presence of a HAS sign ( ρ = 0.632, p = 0.002**). There is also a significant positive correlation between the presence of fibrin-rich and the absence of a HAS sign ( ρ = 0.600, p = 0.004**). Conclusion: Machine learning software can accurately quantify AIS clot composition. The Hounsfield unit density of an AIS clot on a non-contrast CT correlates with the composition of the clot. Measuring the HU of the clot on CT prior to endovascular intervention may inform the clinician of the clot type and help to determine the most appropriate treatment strategy for that clot.