Intraobserver and interobserver variability in schemes for estimating volume of brain lesions on MR images in multiple sclerosis

Abstract

Purpose: Our goal was to evaluate the intraobserver and interobserver reproducibility of measurements of brain lesion load in multiple sclerosis (MS) by using two proposed acquisition schemes.

Methods: Three-millimeter-thick conventional spin-echo (CSE) and fast fluid-attenuated inversion-recovery (FLAIR) sequences were obtained and the lesions segmented using a semiautomated technique based on local thresholding to calculate intraobserver and interobserver reproducibility. These were compared with images obtained from two separate MR units in which 5-mm CSE sequences were obtained and segmented by using the local thresholding technique and also by manual outlining.

Results: The intraobserver coefficient of variation was 4.0% (95% confidence interval [CI], 3.0% to 4.5%) for the 5-mm CSE sequence measured with manual outlining, 3.1% (95% CI, 2.5% to 3.2%) and 5.1% (95% CI, 4.1% to 5.6%) for the two sets of 5-mm CSE sequences measured using the local thresholding technique, 5.7% (95% CI, 3.9% to 6.6%) for the 3-mm CSE sequence, and 2.6% (95% CI, 2.1% to 2.7%) for the fast FLAIR sequence. The interobserver coefficient of variation was 7.1% (95% CI, 4.9% to 8.7%) and 8.3% (95% CI, 6.4% to 9.6%) for the two sets of 5-mm CSE sequences, 7.3% (95% CI, 4.7% to 9.1%) for the 3-mm CSE sequence, and 2.9% (95% CI, 2.3% to 3.3%) for the fast FLAIR sequence. The intraobserver and interobserver reproducibility of measurements obtained with the fast FLAIR technique was significantly better than those obtained with the other techniques.

Conclusions: Our data indicate that the intraobserver and interobserver variability in quantifying MS lesions can be reduced significantly with the use of fast FLAIR sequences, while no significant improvement is gained by reducing the section thickness from 5 mm to 3 mm.