Aurora Semerano, MD
@semerano_aurora
Tools for predicting the success or the failure of reperfusion treatments in the acute setting of ischemic stroke are useful both to assist treatment decision-making and to guide the selection of the best device and reperfusion strategy. Multiple biomarkers and models, including clinical, biochemical, and radiological parameters, are currently under investigations with this purpose. Recently, multimodal analyses of the occlusive clot are receiving growing interest for the potential predictive value on reperfusion outcomes.
Hofmeister et al.(1) addressed this important issue in their recent article in Stroke. More specifically, the authors aimed at identifying the radiomic features of the occlusive clot on pre-treatment non-contrast CT scan, which may predict both first-attempt successful reperfusion with thromboaspiration (defined by modified Treatment in Cerebral Ischemia, mTICI 2b-3) and the number of maneuvers required to achieve successful reperfusion.
After manual segmentation of the occlusive thrombus on baseline imaging, the authors employed a machine learning model of automatic computation to analyze a wide number of radiomic features related to thrombus density, shape, and size. In a retrospective discovery cohort of 109 stroke patients treated with endovascular thrombectomy, the model identified a small subset of radiomic features which were best associated with first-attempt reperfusion. These features resulted to be represented by a lower thrombus density, a bigger variance in thrombus density, and a more homogenous and finer clot texture.
Ultimately, they tested the selected parameters in a prospective validation cohort of 47 patients, to assess their predictive value. Indeed, the selected subset of radiomic features could be able to predict first-attempt successful reperfusion with thromboaspiration with an overall accuracy of 85.1%, a sensitivity of 50.0%, a specificity of 97.1%, and with a receiver operating characteristic curve-AUC of 0.88. The number of maneuvers required to achieve successful reperfusion was predicted with an explained variance of 0.70, a mean squared error of 0.76, and a Pearson correlation coefficient of 0.73.
This study has an interesting exploratory value of the potentiality of clot-related radiomics, which appears a promising technology for treatment personalization in stroke medicine. In this study, measures of thrombus length and shape were not found to be associated with reperfusion outcomes; however, only patients with occlusion of the middle cerebral artery were included. In addition, all patients were treated with a direct aspiration first pass technique (ADAPT) approach. Thus, in future studies, it would be interesting to investigate this radiomic analysis in clots with different occlusion sites and retrieved with different technique approaches. Moreover, it could be hypothesized that the administration of thrombolytic treatment could have at least partially affected thrombus composition after baseline imaging, and future explorations taking into account thrombolytic treatment in adjusted analyses are welcome.
Importantly, the different radiomic features of cerebral thrombi might correspond to different thrombus compositions, which, in turn, may account for differences in retrieval feasibility by endovascular thrombectomy. Recently, the histological analysis of the retrieved cerebral thrombi has been proposed as a tool to optimize the development of thrombectomy devices and retrieval strategies.(2) For example, it has been reported that fibrin-rich thrombi with lower red blood cell content are significantly associated with longer intervention times,(3) whereas controversial results exist about the association with reperfusion outcomes. One study(4) reported higher proportions of red blood cells in thrombi retrieved from patients with successful reperfusion (mTICI 2b-3). Since thrombi with higher red blood cell content are widely recognized as hyperdense on CT scan,(5) this result would be at least indirectly and partially in contrast with the findings of Hofmeister et al. Multidisciplinary efforts are pivotal to advance research in this field. Indeed, studies considering 1) analysis of thrombus composition, 2) advanced imaging tools for prediction of thrombus composition, 3) association of histology and imaging features with reperfusion outcomes and procedure details, and 4) clot-device interactions by means of engineered clot analogs are all warranted to suggest the most suitable, personalized treatment approach in the acute setting of ischemic stroke.
References:
1 Hofmeister J, Bernava G, Rosi A, et al. Clot-Based Radiomics Predict a Mechanical Thrombectomy Strategy for Successful Recanalization in Acute Ischemic Stroke. Stroke. 2020.
2 Bacigaluppi M, Semerano A, Gullotta GS, Strambo D. Insights from thrombi retrieved in stroke due to large vessel occlusion. J of Cerebral Blood Flow and Metabolism. 2019;39:1433-1451.
3 Sporns PB, Hanning U, Schwindt W, et al. Ischemic stroke: histological thrombus composition and pre-interventional CT attenuation are associated with intervention time and rate of secondary embolism. Cerebrovascular Diseases. 2017; 44:344–350.
4 Hashimoto T, Hayakawa M, Funatsu N, et al. Histopathologic analysis of retrieved thrombi associated with successful reperfusion after acute stroke thrombectomy. Stroke. 2016;47:3035–3037.
5 Brinjikji W, Duffy S, Burrows A, et al. Correlation of imaging and histopathology of thrombi in acute ischemic stroke with etiology and outcome: a systematic review. J Neurointerv Surg. 2017;9:529–534.
