Romain Bourcier1, Lili Détraz1, Jean Michel Serfaty2,3, Beatrice Guyomarch Delasalle4, Mahmood Mirza5, Imad Derraz6, Frédérique Toulgoat7, Olivier Naggara8,9, Claire Toquet10, Hubert Desal1. 1. Department of Neuroradiology, CHU de Nantes, Hôpital René et Guillaume Laennec, Saint Herblain, France. 2. CHU de Nantes, Hôpital René et Guillaume Laennec, Imagerie Cardiaque et Vasculaire, Nantes, France. 3. Institut Nationale de la Santé et de la Recherche Médicale, UMR 1087, l'institut du thorax, Paris, France. 4. CHU de Nantes, Hôpital René et Guillaume Laennec, L'institut du thorax, INSERM, CNRS, UNIV Nantes, Nantes, France. 5. Neuravi Thromboembolic Initiative (NTI). 6. Department of Neuroradiology, CHU de Nancy, Nancy, France. 7. Central Department of Radiology and Medical Imaging, CHU de Nantes, Hôtel Dieu, Nantes, France. 8. Department of Neuroradiology, Centre Hospitalier Sainte-Anne. 9. Université Paris-Descartes, INSERM U 894, Paris, France. 10. Department of Pathology, CHU de Nantes, Hôtel Dieu, Nantes, France.
Abstract
BACKGROUND AND PURPOSE: The susceptibility vessel sign (SVS) on magnetic resonance imaging (MRI) is related to thrombus location, composition, and size in acute stroke. No previous study has determined its inter-MRI scanner variability. We aimed to compare the diagnostic accuracy in-vitro of four different MRI scanners for the characterization of histologic thrombus composition. METHODS: Thirty-five manufactured thrombi analogs of different composition that were histologically categorized as fibrin-dominant, mixed, or red blood cell (RBC)-dominant were scanned on four different MRI units with T2* sequence. Nine radiologists, blinded to thrombus composition and MRI scanner model, classified twice, in a 2-week interval, the SVS of each thrombus as absent, questionable, or present. We calculated the weighted kappa with 95% confidence interval (CI), sensitivity, specificity and accuracy of the SVS on each MRI scanner to detect RBC-dominant thrombi. RESULTS: The SVS was present in 42%, absent in 33%, and questionable in 25% of thrombi. The interscanner agreement was moderate to good, ranging from .45 (CI: .37-.52) to .67 (CI: .61-.74). The correlation between the SVS and the thrombus composition was moderate (κ: .50 [CI: .44-.55]) to good κ: .76 ([CI: .72-.80]). Sensitivity, specificity, and accuracy to identify RBC-dominant clots were significantly different between MRI scanners (P < .001). CONCLUSION: The diagnostic accuracy of SVS to determine thrombus composition varies significantly among MRI scanners. Normalization of T2*sequences between scanners may be needed to better predict thrombus composition in multicenter studies.
BACKGROUND AND PURPOSE: The susceptibility vessel sign (SVS) on magnetic resonance imaging (MRI) is related to thrombus location, composition, and size in acute stroke. No previous study has determined its inter-MRI scanner variability. We aimed to compare the diagnostic accuracy in-vitro of four different MRI scanners for the characterization of histologic thrombus composition. METHODS: Thirty-five manufactured thrombi analogs of different composition that were histologically categorized as fibrin-dominant, mixed, or red blood cell (RBC)-dominant were scanned on four different MRI units with T2* sequence. Nine radiologists, blinded to thrombus composition and MRI scanner model, classified twice, in a 2-week interval, the SVS of each thrombus as absent, questionable, or present. We calculated the weighted kappa with 95% confidence interval (CI), sensitivity, specificity and accuracy of the SVS on each MRI scanner to detect RBC-dominant thrombi. RESULTS: The SVS was present in 42%, absent in 33%, and questionable in 25% of thrombi. The interscanner agreement was moderate to good, ranging from .45 (CI: .37-.52) to .67 (CI: .61-.74). The correlation between the SVS and the thrombus composition was moderate (κ: .50 [CI: .44-.55]) to good κ: .76 ([CI: .72-.80]). Sensitivity, specificity, and accuracy to identify RBC-dominant clots were significantly different between MRI scanners (P < .001). CONCLUSION: The diagnostic accuracy of SVS to determine thrombus composition varies significantly among MRI scanners. Normalization of T2*sequences between scanners may be needed to better predict thrombus composition in multicenter studies.
Authors: Waleed Brinjikji; Sharon Duffy; Anthony Burrows; Werner Hacke; David Liebeskind; Charles B L M Majoie; Diederik W J Dippel; Adnan H Siddiqui; Pooja Khatri; Blaise Baxter; Raul Nogeuira; Matt Gounis; Tudor Jovin; David F Kallmes Journal: J Neurointerv Surg Date: 2016-05-10 Impact factor: 5.836
Authors: R Bourcier; R Pautre; M Mirza; C Castets; J Darcourt; J Labreuche; L Detraz; H Desal; J-M Serfaty; C Toquet Journal: AJNR Am J Neuroradiol Date: 2019-01 Impact factor: 3.825
Authors: I Derraz; M Pou; J Labreuche; L Legrand; S Soize; M Tisserand; C Rosso; M Piotin; G Boulouis; C Oppenheim; O Naggara; S Bracard; F Clarençon; B Lapergue; R Bourcier Journal: AJNR Am J Neuroradiol Date: 2020-11-12 Impact factor: 3.825