Intercenter agreement of brain atrophy measurement in multiple sclerosis patients using manually-edited SIENA and SIENAX

J Magn Reson Imaging. 2007 Oct;26(4):881-5. doi: 10.1002/jmri.21101.


Purpose: To investigate intercenter agreement of brain volume (change) measurement in multiple sclerosis (MS) using structural image evaluation using normalization of atrophy (SIENA) and the cross-sectional version of SIENA (SIENAX) with additional manual editing to correct for inadequate brain extraction.

Materials and methods: Baseline and follow-up T1-weighted MR images of 20 MS patients were dispatched to five centers. Each center performed fully-automated and manually-edited analyses for SIENAX, yielding normalized brain volume (NBV), and SIENA, yielding percentage brain volume change (PBVC). Intercenter agreement was assessed with the concordance correlation coefficient (CCC).

Results: Intercenter agreement was perfect for fully automated NBV and PBVC (both CCC = 1.0), and remained substantial upon manual editing (CCC = 0.94 for NBV, CCC = 0.95 for PBVC). Mean NBV values for each center decreased significantly after manual editing (overall mean NBV = 1605.3 cm(3) vs. 1651.1 cm(3) without manual editing; t = -4.58, P < 0.001). Total variance in PBVC decreased significantly by a factor of 1.8 after manual editing (sigma(2) = 2.82 before, and sigma(2) = 1.54 after manual editing, P < 0.05).

Conclusion: Substantial intercenter agreement was found for manually-edited SIENAX and SIENA, suggesting that measurements from multiple centers may be pooled. Manual editing reduces overestimation of NBV, and is likely to increase statistical power for PBVC.

Publication types

  • Multicenter Study
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Atrophy*
  • Automation
  • Brain / metabolism
  • Brain / pathology*
  • Female
  • Humans
  • Image Processing, Computer-Assisted / methods*
  • Magnetic Resonance Imaging / methods*
  • Male
  • Models, Statistical
  • Multiple Sclerosis / pathology*
  • Pattern Recognition, Automated
  • Reproducibility of Results