Abstract W P30: Implementation of a Web-based Approximate Volume Tool for Acute Stroke Lesion Measurement Using MRI

Abstract

Background and Purpose: In the clinical setting of acute stroke, determination of the amount of penumbra quickly is increasingly important for treatment decisions. The approximate ABC/2 method is an alternative when image quality restricts the usage of automated methods. Our hypothesis is that usage of a web-based ABC/2 app will increase the predictive accuracy of the users’ visual and approximate mismatch estimates. Methods: Ten training cases were selected retrospectively from patients diagnosed with acute ischemic stroke at Suburban Hospital or MedStar Washington Hospital Center and scanned acutely with MRI between January 1, 2014 through February 28, 2014. Image readers including four board-certified vascular neurologists and two vascular neurology fellows used DICOM software to initially perform DWI and MTT visual estimates and then to calculate the ABC/2 volumes for DWI, MTT and mismatch using a web based tool. After one month, the vascular neurologists were re-tested. The results from the initial training and re-testing were compared to calculate the agreement of visual mismatch classification versus the planimetric volume calculations, where mismatch was defined as MTT- DWI ≥50 ml across sessions. The correlations of approximate and planimetric calculations were compared for the DWI and MTT measurements. Results: For the DWI approximate measurements from the training to the re-testing, the Spearman correlation coefficients with the planimetric measurements increased for four out of the six vascular neurologists with r =[0.78 to 0.96], [0.81-0.88], [0.65-0.94] and [0.88-0.98]. Two readers demonstrated highly correlated DWI measurements, [0.98-0.96] and [0.99-0.93]. For the MTT, the correlations increased for three out of the six vascular neurologists with training to re-testing r =[0.76 to 0.93], [0.87-0.88] and [0.56-0.92]. Two readers demonstrated comparable MTT measurements, [0.94-0.93] and [0.92-0.87]. One reader’s MTT correlation decreased from [0.89-0.71]. The average agreement of the visual mismatch estimates increased from 63% to 70%. Conclusions: Using a web-based ABC/2 app was beneficial for increasing the visual mismatch agreement and accuracy of approximate measurements for multiple vascular neurologists.