Quantification of multiple sclerosis lesion volumes in 1.5 and 0.5 T anisotropically filtered and unfiltered MR exams

Recently, guidelines for the use of MRI in the monitoring of MS have recommended the use of imaging systems with mid-field (0.5-1.0 T) or high- field (greater than 1.0 T) strengths. Higher field strengths provide many advantages, including increased signal-to-noise ratios (SNR). SNR also may be increased by post-processing algorithms that reduce noise. In this paper we evaluate the impact on operator variability of (a) lesion quantification in high-field (1.5 T) versus mid-field (0.5 T) exams; and (b) an anisotropic diffusion filter algorithm that reduces image noise without blurring or moving object boundaries. Inter- and intra-operator reliability and variability were studied using repeated quantification of lesions in 1.5 and 0.5 T filtered and unfiltered MR exams of a MS patient. Results indicate that inter-operator variability in 1.5 T unfiltered exams was 0.34 cm³ and was significantly larger than that in 1.5 T filtered (0.27 cm³), 0.5 T unfiltered (0.26 cm³), and 0.5 T filtered (0.24 cm³) exams. Similarly, intra-operator variability in 1.5 T unfiltered exams was 0.23 cm³ and was significantly larger than that in 1.5 T filtered (0.19 cm³), 0.5 T unfiltered (0.19 cm³), and 0.5 T filtered (0.18 cm³) exams. In addition, the minimum significant change between two successive measurements of lesion volume by the same operator, was 0.64 cm³ in 1.5 T unfiltered exams, but 0.53 cm³ or less in other exams. For two different operators making successive measurements, the minimum significant change was 0.94 cm³ in 1.5 T unfiltered exams, but only 0.75 cm³ or less in other exams. Finally, the number of lesions to be monitored for an average change in volume at a given power and significance level was greater by 30%-60% for quantification in 1.5 T unfiltered exams. These results suggest that inter- and intra-operator variability are reduced by anisotropic filtering, and by quantification in 0.5 T exams. Reduced operator variabilities may result from higher detail signal-to-noise ratios (dSNRs) in 0.5 T and filtered exams.