Comparing magnetic resonance imaging brain atrophy measurement techniques in multiple sclerosis clinical practice: longitudinal assessment

    loading  Checking for direct PDF access through Ovid



To determine the annualised percentage whole brain volume change (PBVC) between two time-points using Structural Image Evaluation using Normalisation of Atrophy (SIENA), SIENA Cross-sectional (SIENAX), and MSmetrixTM-longitudinal (MSmetrix-long) in a real-world multiple sclerosis (MS) cohort. To compare the whole brain atrophy (WBA) measurements derived by two registration-based longitudinal software methods (SIENA, MSmetrix-long), and a cross-sectional, segmentation-based method (SIENAX).


Clinical MRI scans (1 mm3 3DT1 IR-FSPGR; GE 3.0T scanner) were acquired from 102 MS patients at baseline and approximately 12 months. WBA analysis was performed using SIENA, SIENAX and MSmetrix-long (fully automated) for each subject. Annualised PBVC was calculated using all methods, and correlations between the measurement techniques were evaluated using linear regression.


Mean (SD) annualised PBVC estimated by SIENA, SIENAX and MSmetrix-long was −0.64% (0.73) −0.49% (3.05) and −0.59% (0.65), respectively. Annualised PBVC was >0.4% loss in 55.88% of subjects using SIENA, 56.86% using SIENAX and 57.84% using MSmetrix-long. There was a strong correlation (Pearson’s r) between SIENA and MSmetrix-long derived measures of annualised PBVC; 0.805 (p<0.001). Correlations between SIENAX and the other methods were weaker; 0.313 compared with SIENA, and 0.264 compared with MSmetrix-long (p<0.01).


WBA measurements derived using the fully automated MSmetrix-long correlated strongly with the results using SIENA, however, measurements using SIENAX correlated weakly. Longitudinal registration-based techniques hold promise for use in MS clinical practice but are currently not validated at the individual patient level. Fully automated, easy-to-use platforms are likely to accelerate translation from the research setting into routine MS care, and subsequently warrant further investigation. Segmentation-based methods currently have measurement error rates that render them suboptimal for use in longitudinal assessment, especially at the individual level. Technical improvements and the development of reference populations may result in these techniques having a place in future MS patient care.

Related Topics

    loading  Loading Related Articles