6533b85bfe1ef96bd12ba1c0

RESEARCH PRODUCT

Multicenter stability of diffusion tensor imaging measures: a European clinical and physical phantom study.

Stefan J. TeipelStefan J. TeipelMassimo FilippiFrank JessenAndreas FellgiebelBram StieltjesKarl Heinz HauensteinFrank HentschelPetra J. W. PouwelsThomas MeindlUlrike ErnemannStefan KlöppelHarald HampelJulio Acosta-cabroneroGiovanni B. FrisoniSigrid ReuterSigrid Reuter

subject

Coefficient of variationNeuroscience (miscellaneous)Nerve Fibers MyelinatedBrain mappingImaging phantommethods [Diffusion Tensor Imaging]White matterYoung AdultNeuroimagingBiasAlzheimer Diseasepathology [Brain]Fractional anisotropymedicineHumansRadiology Nuclear Medicine and imagingddc:610AgedAged 80 and overReproducibilityBrain Mappingpathology [Nerve Fibers Myelinated]business.industryPhantoms Imagingdiagnosis [Alzheimer Disease]BrainMiddle AgedEuropePsychiatry and Mental healthDiffusion Tensor Imagingmedicine.anatomical_structureAnisotropyFemaleNuclear medicinebusinessPsychologyDiffusion MRI

description

Diffusion tensor imaging (DTI) detects white matter damage in neuro-psychiatric disorders, but data on reliability of DTI measures across more than two scanners are still missing. In this study we assessed multicenter reproducibility of DTI acquisitions based on a physical phantom as well as brain scans across 16 scanners. In addition, we performed DTI scans in a group of 26 patients with clinically probable Alzheimer's disease (AD) and 12 healthy elderly controls at one single center. We determined the variability of fractional anisotropy (FA) measures using manually placed regions of interest as well as automated tract based spatial statistics and deformation based analysis. The coefficient of variation (CV) of FA was 6.9% for the physical phantom data. The mean CV across the multicenter brain scans was 14% for tract based statistics, and 29% for deformation based analysis. The degree of variation was higher in less organized fiber tracts. Our findings suggest that a clinical and physical phantom study involving more than two scanners is indispensable to detect potential sources of bias and to reliably estimate effect size in multicenter diagnostic trials using DTI.

yearjournalcountryeditionlanguage
2011-01-01
10.1016/j.pscychresns.2011.05.012https://pub.dzne.de/record/136392