Abstract WMP21: Texture Analysis of Intracranial Thrombus Using CT and CTA Predicts Early Recanalization With Intravenous Alteplase

• Published in: Stroke (1970) Vol. 49; no. Suppl_1
• Main Authors: Qiu, Wu, Kuang, Hulin, Najm, Mohamed, McDougall, Connor, Nair, Jay, McDougall, Brooklyn, Chung, Kevin, Wilson, Alexis T, Goyal, Mayank, Hill, Michael, Demchuk, Andrew, Menon, Bijoy K
• Format: Journal Article
• Language: English
• Published: 22.01.2018
• ISSN: 0039-2499; 1524-4628
• DOI: 10.1161/str.49.suppl_1.WMP21

Abstract only Objective: In acute ischemic stroke (AIS), thrombus lysis and early restoration of blood flow to ischemic brain is the goal of reperfusion therapy. We hypothesize that image texture analysis of thrombus using non-contrast CT (NCCT) and CTA can predict early recanalization with IV alteplase in AIS patients with ICA/M1 MCA thrombi. Methods: We used a case control design for the study using the whole cohort of the clinical trial of Identifying New approaches to optimize Thrombus characterization for predicting Early Recanalization and Reperfusion with iv tPA using Serial CT angiography (INTERRSeCT). Cases were 31 patients with ICA/M1 MCA thrombi treated with IV alteplase, who achieved early recanalization (as assessed on first angio run in patients undergoing additional IA therapy or on repeat CTA at 4 hrs). Cases were matched with controls (37 patients with ICA/M1 MCA thrombi treated with IV alteplase who did not achieve recanalization). All patients had thin slice NCCT (slice thickness < 2.5 mm) and CTA (≤ 0.625mm). Thrombi in ICA/M1 MCA region were manually contoured from NCCT guided by co-registered CTA. We extracted 378 features (including thrombus length and volume, first order statistics, and image texture features) from each thrombus in both NCCT and CTA images. This was followed by feature selection using linear discriminative analysis. The top 5 features from NCCT and CTA images along with the clot texture difference between NCCT and CTA were used to train a linear support vector machine classifier. Five times 6-fold cross validation was used to evaluate the trained classifier. Results: Receiver operator curves show that thrombus texture from NCCT and CTA are predictive of early recanalization with IV alteplase. Combined NCCT & CTA features and texture difference is the best predictor (AUC: 0.82). Conclusions: Thrombus texture features from NCCT and CTA are strongly predictive of early recanalization with IV alteplase in AIS patients with proximal occlusions.