Objectives:
In the clinical setting, the extent of mismatch on MRI is frequently assessed with an approximate “XYZ/2” method but the agreement with the “gold standard” planimetric volume and the “visual evaluation” methods are not known. In a published study, we established that the visual evaluation and planimetric methods are equivalent as far as mismatch classification. The objectives of this study were to quantify the agreement of the approximate method with the “gold standard” and “visual evaluation” methods and to compare the mismatch classification results.
Methods:
Patients were selected from the Lesion Evolution of Stroke and Ischemia On Neuroimaging (LESION) database if they: had an acute ischemic stroke, were treated with intravenous rt-PA only, and had a pre-treatment MRI with evaluable maps including trace or isotropic b1000 DWI and MTT images. A trained rater viewed the images on the PACS, placed the two perpendicular linear measurements, “X” and “Y”, on the slices with the largest DWI and MTT lesion areas, and then used a “XYZ/2” formula where “Z” was the product of the slice thickness and the total number of slices containing the lesion. A separate expert rater measured the planimetric volumes on a slice-by-slice basis with a semiautomated segmentation tool followed by manual editing. Expert readers evaluated the MRI scans for the presence of qualitative mismatch. The expert readers were not the trained reader that performed the approximate volume measurements. Quantitative mismatch was considered present if MTT volume - DWI volume ≥50 ml. Mismatch classifications using the ≥ 50 ml definition were compared by constructing contingency tables.
Results:
A total of 194 patients met the study criteria and had median DWI and MTT planimetric volumes of 13.06 ml and 99.27 ml respectively. For both the DWI (n=170) and MTT (n=164), 94% of the measurements were within two standard deviations of the difference between the planimetric and approximate volume measurements. Comparing the planimetric and approximate volume measurements, the Spearman correlation coefficients were 0.855 and 0.886 for the DWI and MTT measurements respectively (p<0.01). Compared to the planimetric method, the approximate “XYZ/2” method had a high sensitivity (0.91), specificity (0.80), accuracy (0.86) and positive predictive value (0.85) to detect mismatch using the ≥ 50 ml definition. Compared to the qualitative method, the approximate “XYZ/2” method had a sensitivity (0.77), specificity (0.76), accuracy (0.77) and positive predictive value (0.87) to detect mismatch using the ≥ 50 ml definition.
Conclusions:
The approximate “XYZ/2” method is sufficient for classifying the presence of MRI determined mismatch in acute stroke patients and therefore is a potential tool when using MRI determined mismatch as an inclusion criteria for clinical trials.