The SIENA/FSL whole brain atrophy algorithm is no more reproducible at 3 T than 1.5 T for Alzheimer's disease

K.S. Cover, R.A. van Schijndel, V. Popescu, A. Redolfi, D.L. Knol, G. B. Frisoni, F. Barkhof, H. Vrenken, B.W. van Dijk

Research output: Contribution to journalArticleAcademicpeer-review

12 Citations (Scopus)


The back-to-back (BTB) acquisition of MP-RAGE MRI scans of the Alzheimer's Disease Neuroimaging Initiative (ADNI1) provides an excellent data set with which to check the reproducibility of brain atrophy measures. As part of ADNI1, 131 subjects received BTB MP-RAGEs at multiple time points and two field strengths of 3. T and 1.5. T. As a result, high quality data from 200 subject-visit-pairs was available to compare the reproducibility of brain atrophies measured with FSL/SIENA over 12 to 18 month intervals at both 3. T and 1.5. T. Although several publications have reported on the differing performance of brain atrophy measures at 3. T and 1.5. T, no formal comparison of reproducibility has been published to date. Another goal was to check whether tuning SIENA options, including -B, -S, -R and the fractional intensity threshold ( f) had a significant impact on the reproducibility. The BTB reproducibility for SIENA was quantified by the 50th percentile of the absolute value of the difference in the percentage brain volume change (PBVC) for the BTB MP-RAGES. At both 3. T and 1.5. T the SIENA option combination of "-B f=0.2", which is different from the default values of f=0.5, yielded the best reproducibility as measured by the 50th percentile yielding 0.28 (0.23-0.39)% and 0.26 (0.20-0.32)%. These results demonstrated that in general 3. T had no advantage over 1.5. T for the whole brain atrophy measure - at least for SIENA. While 3. T MRI is superior to 1.5. T for many types of measurements, and thus worth the additional cost, brain atrophy measurement does not seem to be one of them. © 2014 Elsevier Ireland Ltd.
Original languageEnglish
Pages (from-to)14-21
JournalPsychiatry Research: Neuroimaging
Issue number1
Publication statusPublished - 2014

Cite this