|| Checking for direct PDF access through Ovid
Background: The presence of a perfusion deficit in an acute stroke patient can play an important role in their clinical management. However, many patients are unable to have perfusion-weighted imaging (PWI) due to renal disease. Perfusion deficits are often accompanied by FLAIR hyperintense vessels (FHV), presumably due to slow arterial blood flow, and GRE hypointense vessels (GHV), presumably due to venous congestion.Purpose: To determine how well FHV and GHV perform at identifying PWI lesions.Methods: One rater, blinded to the PWI MR sequences, retrospectively reviewed the DWI, FLAIR and GRE scans of acute stroke patients enrolled in the NIH Natural History of Stroke study during 2013-2014 who had an MRI with PWI prior to being treated with IV tPA. DWI images were used to guide evaluation of the FLAIR and GRE images for FHV and GHV and in each case were classified as definitively present, possibly present or absent. PWI lesion volumes were calculated by thresholding the time-to-peak (TTP) maps at 4 seconds beyond normal tissue. ROC analysis was used to assess the performance of FHV and GHV at various PWI lesion volume thresholds.Results: 102 patients were included in the analysis; their mean PWI lesion volume was 52 mL with a standard deviation of +/- 66 mL. 22% of patients had no perfusion deficit. The ROC analysis found the presence of any FHV performed the best with an area under the curve (AUC) of 0.925 displayed in the figure. Any GHV performed modestly with an AUC of 0.776. Combining FHV with GHV did not improve the performance over FHV alone (AUC=0.876). The sensitivity and specificity for identifying any perfusion deficit with FHV was 95% and 67% respectively with 87% being correctly classified. For detecting a PWI lesion greater than 10 mL, FHV had an 80% sensitivity and 93% specificity classifying 83% correctly.Conclusions: FHV is highly sensitive for identifying a perfusion deficit in stroke patients, and for patients with a lesion volume greater than 10 mL it is highly specific.