Diagnostic performance of intraoperative cone beam computed tomography compared with postoperative magnetic resonance imaging for detecting hemorrhagic transformation after endovascular treatment following large vessel occlusion.

ABSTRACT

OBJECTIVES: Early detection of hemorrhagic transformation (HT) in patients with large vessel occlusion (LVO) after endovascular treatment is important for postoperative patient management. We investigated the diagnostic performance of intraoperative cone beam computed tomography (CBCT) with reference standard magnetic resonance imaging (MRI) for detecting HT. MATERIALS AND METHODS: Consecutive patients with LVO treated by endovascular treatment who underwent intraoperative CBCT and postoperative MRI were included. Two observers evaluated all images for the presence of HT. Sensitivity and specificity for detecting HT were calculated with MRI as reference standard. The observers classified HT according to the European Cooperative Acute Stroke Study (ECASS). Inter-method and inter-rater agreement for the detection of HT and for the ECASS classification were assessed using kappa or weighted Brennan-Prediger (wBP) statistics. RESULTS: Images of 106 procedures (94 for anterior circulation) were analyzed. The sensitivity and specificity for detecting HT on CBCT were 0.77 and 0.83, respectively, for all procedures and 0.83 and 0.8, respectively, for anterior circulation. The inter-method agreement for HT detection (κ = 0.63 overall, κ = 0.69 anterior circulation) and ECASS classification (wBP = 0.67 overall, wBP = 0.77 anterior circulation) were substantial. The inter-rater agreement for HT detection (κ = 0.87 overall, κ = 0.85 anterior circulation) and for ECASS classification (wBP = 0.95 overall, wBP = 0.92 anterior circulation) were almost perfect. CONCLUSIONS: The diagnostic performance of CBCT for the detection of HT in stroke patients treated for LVO was acceptable with excellent inter-rater agreement. Intraoperative CBCT may be useful to trigger early interventions if HT is detected, although detailed classifications of HT may be difficult.