Thrombus radiomics in patients with anterior circulation acute ischemic stroke undergoing endovascular treatment

Abstract

BackgroundThrombus radiomics (TR) describe complex shape and textural thrombus imaging features. We aimed to study the relationship of TR extracted from non-contrast CT with procedural and functional outcome in endovascular-treated patients with acute ischemic stroke.MethodsThrombi were segmented on thin-slice non-contrast CT (≤1 mm) from 699 patients included in the MR CLEAN Registry. In a pilot study, we selected 51 TR with consistent values across two raters’ segmentations (ICC >0.75). Random forest models using TR in addition or as a substitute to baseline clinical variables (CV) and manual thrombus measurements (MTM) were trained with 499 patients and evaluated on 200 patients for predicting successful reperfusion (extended Thrombolysis in Cerebral Ischemia (eTICI) ≥2B), first attempt reperfusion, reperfusion within three attempts, and functional independence (modified Rankin Scale (mRS) ≤2). Three texture and shape features were selected based on feature importance and related to eTICI ≥2B, number of attempts to eTICI ≥2B, and 90-day mRS with ordinal logistic regression.ResultsRandom forest models using TR, CV or MTM had comparable predictive performance. Thrombus texture (inverse difference moment normalized) was independently associated with reperfusion (adjusted common OR (acOR) 0.85, 95% CI 0.72 to 0.99). Thrombus volume and texture were also independently associated with the number of attempts to successful reperfusion (acOR 1.36, 95% CI 1.03 to 1.88 and acOR 1.24, 95% CI 1.04 to 1.49).ConclusionsTR describing thrombus volume and texture were associated with more attempts to successful reperfusion. Compared with models using CV and MTM, TR had no added value for predicting procedural and functional outcome.

Radiomics-based intracranial thrombus features on preoperative noncontrast CT predicts successful recanalization of mechanical thrombectomy in acute ischemic stroke

“…Another paper utilizes single-phase CTA for thrombus measurements and explores the efficiency of thrombus texture to predict more attempts to successful reperfusion in patients with anterior circulation stroke. These thrombus texture, however, were not related to functional outcome ( 27 ). Thus, we formulate a hypothesis that the biochemical characteristics of the thrombus can contributed to decision-making and treatment planning in patients receiving SR.…”

Predictive value of clot imaging in acute ischemic stroke: A systematic review of artificial intelligence and conventional studies

“…Because of the unavailability of automatic segmentation methods, and because the manual segmentation of the clot is a tedious process, conventional studies resort so far to manual delineation of ROIs, in CT scans slices, to read the density values and to provide image markers associated with clot composition and organization, such as the HAS and perviousness measures. Disposing of standardized means, across multiple centers, of gathering comprehensive information about the clot through 3D renderings could not only provide accurate, meaningful variables for conventional studies, but also accelerate AI developments [17]. In perspective, clot characteristics obtained from automatic 3D renderings could be used as variables for training AI models predictive for clot composition and response to treatment: along with measures of density and perviousness, the clot extent and shape, together with its location, can provide a wealth of information.…”

“…The study of Hofmeister et al [15] utilizes a support vector machine classifier to train and validate a model predicting first-attempt recanalization with thromboaspiration, and a support vector regression to train and validate a model predicting the number of passages required for successful recanalization with stent retriever. The study of Sarioglu et al [16] utilizes the selected RFs as variables in a logistic regression model, along with demographic and clinical variables, and van Voorst et al [17] compares the predictive performance of random forest models using either a set of 6 thrombus radiomic features, a set of 19 clinical variables, a set of 3 manually extracted thrombus measurements (density, perviousness, thrombus length), or combinations of these sets of variables.…”

“…The procedures for selecting the relevant RFs, which are used in model development, differ vastly among the studies, and the number of RFs identified as predictive is different. Not surprisingly, for studies considering additional predictor variables, along with RFs, the findings are also very different: Qiu et al found RFs more predictive than conventional CT derived measures of thrombus [14], Sarioglu et al found that the diagnostic accuracy is increased when combining RFs with demographic and clinical variables [16], and van Voorst et al [17] found that random forest models using thrombus radiomics, clinical variables or manual thrombus measurements have comparable predictive performance.…”

“…Some are single-center studies [13] [15] [16], or use a small sample size [14], and merely establish a proof-ofconcept for models predictive for thrombus characteristics or various treatment-related outcomes. Compared to the other studies, van Voorst et al [17] is using a larger data set while extracting a relatively low number of RFs per thrombus. The generalizability of his findings is hampered by the lack of comparison with radiomics models evaluating different sets of RFs.…”