Reproducibility of brain MRI lesion volume measurements in multiple sclerosis using a local thresholding technique: effects of formal operator training.

The assessment of lesion load (LL) on brain magnetic resonance imaging (MRI) scans from patients with multiple sclerosis (MS) is widely used to monitor disease evolution, natural or modified by treatments. In this study, we evaluated the effect of formal operator training on the intra- and inter-observer reproducibility of LL measurements obtained by several operators in a setting similar to that of clinical trials. Proton-density (PD)-weighted, unenhanced and enhanced T1-weighted brain MRI scans were obtained from 10 MS patients. Five naive technicians assessed LL on these images, using a semiautomated local thresholding technique for lesion segmentation and marked hardcopies as a reference. Measurements were performed twice before and twice after a 20-hour operator training. Mean intra-observer measurement coefficient of variations (COV) before and after the training were 3.1 and 1.6% for PD-weighted LL, 4.3 and 1.8% for unenhanced T1-weighted LL (p < 0. 001), 4.9 and 2.0% for enhanced T1-weighted LL (p = 0.002). Mean inter-observer COV were significantly reduced after training (from 10.0 to 5.6% for PD-weighted, from 11.0 to 7.3% for unenhanced T1-weighted and from 16.0 to 6.8% for enhanced T1-weighted LL). Our data indicate that LL assessment on serial MRI scans from MS patients performed by technicians, using a local thresholding technique for lesion segmentation, is characterized by low measurement variability which may be significantly improved by a short and cost-effective training.