Abstract TP68: Visual Assessment of Diffusion-Weighted-Imaging Infarct Volume Lacks Accuracy and Reliability

Abstract

Introduction: Patients screened by MRI in the extended window (6-24hr) for mechanical thrombectomy (MT) might require the treating physician to visually estimate the DWI infarct volume at institutions where the RAPID software is not available. We aimed to assess the accuracy of DWI volume estimation compared with RAPID, as well as the interrater and intrarater agreement among physicians. Methods: Eighteen raters (all physicians involved on daily stroke management) assessed over two independent sessions 32 MR scans from a prospective MT database. Raters generated a visual estimate of the DWI infarct volume and were asked whether they would recommend MT for the assessed patient or not. Sensitivity, specificity and accuracy of the estimated volumes were compared with the available RAPID measurements for various volume cut-points. Interrater and intrarater agreement was measured using Fleiss’ kappa statistics. Results: Raters mean accuracy was <90% for all volume cut-points ( A ). Interrater agreement was below substantial for each DWI infarct volume cut-point ( B ). Intrarater agreement was substantial for 55-83% of raters, depending on the selected cut-point ( C ). Raters recommended MT for a mean of 77% of patients, including a mean of 56% of patients with an estimated infarct volume≥71ml. Conclusion: The visual assessment of DWI infarct volume lacks accuracy when RAPID measurements are the reference values. Less-than-substantial reproducibility of the volume estimates was demonstrated among physician raters who recommended MT for large infarct volumes despite the absence of rigorous scientific evidence.