The purpose of this study was to compare measures of lesion volume obtained by means of 1.5-T MR imaging to those obtained by the Cavalieri method, 6 weeks after experimental spinal cord injury.
Nine male Wistar rats were subjected to spinal cord injury by clip compression (50 g) at the T-4 level. Six weeks postinjury, the rats were sacrificed, and spinal cords were analyzed ex vivo for lesion volume by means of 1.5-T MR imaging and subsequently, by the Cavalieri method. In the latter method, cords were cut longitudinally in 25-μm sections and stained with solochrome cyanin for myelin. The area of the lesion was determined for each serial section, and the distance-weighted sum of all area measures was then calculated to estimate the total lesion volume.
Bland–Altman analysis showed that the 2 methods had an acceptable level of agreement for lesion volume estimation, but the Cavalieri method was prone to an overestimation bias. The MR imaging estimates of lesion volume were greater than the Cavalieri method estimates in 3 spinal cords, but the difference between measures was within 1 standard deviation of perfect agreement in these 3 lesions, and the mean difference between measures was 18.3%. In contrast, in those lesions in which the Cavalieri method yielded larger lesion volumes (5 lesions), the difference between measures was 2 standard deviations away from perfect agreement for 2 animals and the mean difference between measures was 72.4%.
The results illustrate that the overestimation bias of the Cavalieri method is due, in part, to artifacts produced during processing of the spinal cord tissue.