Computed tomography perfusion in stroke mimics

Abstract

Objective To describe the prevalence and patterns of abnormal findings on automated computed tomography perfusion in patients with stroke mimic. Methods We reviewed a retrospective multi-site cohort of consecutive patients undergoing computed tomography perfusion for suspected acute ischemic stroke within 24 h from last normal (June 2017 to December 2017). The primary outcome was the diagnosis of stroke mimic. Hypoperfusion abnormalities on iSchemaView RAPID automated computed tomography perfusion were compared between patients with stroke/transient ischemic attack and stroke mimic using mixed-effects multivariable logistic regression, focusing on absence of perfusion abnormalities and discordance with clinical symptoms and computed tomography angiography findings. Results Of 410 consecutive patients who underwent computed tomography perfusion, 348 met inclusion criteria (178 (51%) stroke, 19 (6%) transient ischemic attack, and 151 (43%) mimic). Time-to-maximum of the tissue residue function (Tmax>6s) abnormalities were seen in 42 (28%) patients with stroke mimic and 122 (62%) patients with stroke/transient ischemic attack ( p < 0.001). Patients with stroke mimic were more likely to have a normal Tmax pattern (volume = 0mL; adjusted OR: 2.2, 95% CI: 1.1–4.3, p = 0.02). When the Tmax pattern was abnormal, a higher proportion of patients with stroke mimic had Tmax patterns fully discordant with clinical symptoms than patients with stroke/transient ischemic attack (28/39 (71%) vs. 10/115 (9%), p < 0.001). Fully discordant Tmax abnormalities were strongly associated with stroke mimic (adjusted OR: 48.6, 95% CI: 7.0–336, p < 0.001), with a negative predictive value for identifying mimic of 91% (95% CI: 85–94%). Conclusion While one-quarter of patients with stroke mimic show Tmax abnormalities on automated RAPID computed tomography perfusion imaging, the majority of patterns were discordant with symptoms and vessel status. Normal or fully discordant Tmax abnormalities are were more common with stroke mimic and may inform stroke treatment decision making.