Interrater Agreement of Computed Tomography Infarct Measurement

Publisher Website
Google Scholar
Abstract
To determine the reliability of infarct measurements on hardcopy computed tomography (CT) images the in vivo (IV) infarct volumes of 20 CT‐detected infarcts were estimated and divided into four size groups with 5 infarcts in each group: Group A, less than 0.5 ml; Group 8, 0.5 to 5.0 ml; Group C, 5 to 50 ml; and Group D, more than 50 mi. Seventeen infarcts were measured once and 3 infarcts three times to the nearest 0.5 mm by each of two neurologists and two neuroradiologists using a ruler on hard‐copy CT images. The longest diameter (designated AP), the greatest diameter at right angles to AP (designated LAT), and the number of slices showing the infarct were recorded and multiplied by the hard‐copy minification factor to give IV dimensions. Volume (VOL) was calculated according to a previously published method. lnterrater intraclass correlation coefficients for all infarcts combined were 0.98 (AP), 0.91 (LAT), and 0.97 (VOL). Using all raters' measurements for any single infarct, the difference between the largest and the smallest measurement of AP and LAT was smallest (< 6 mm IV) for Group A and largest (< 31 mm IV) for Group D. This difference was largest relative to the dimension being measured in Groups A and 8, where it reached 101% of the mean of the four raters' measurements for the AP dimension being measured in Group 8, and 70% of the dimension being measured in Group A. With all raters' measurements for any single infarct, the difference between the largest and smallest measurement of VOL was smallest (< 0.5 ml) for Group A and largest (< 260 ml) for Group D. This difference was largest relative to the mean volume of the infarct being measured in Group 8, where it reached 153% of the mean of the four raters' measurements for VOL and reached 115% of the mean of the four raters' measurements for VOL in Group A. The authors conclude that infarcts can be measured on hard‐copy images with good interrater agreement. When infarcts with a volume smaller than 5 ml are measured, differences between raters' measurements may exceed the size of the dimensions being measured.