Basic MR sequence parameters systematically bias automated brain volume estimation.

Details

Serval ID
serval:BIB_F4133A1D01DD
Type
Article: article from journal or magazin.
Collection
Publications
Institution
Title
Basic MR sequence parameters systematically bias automated brain volume estimation.
Journal
Neuroradiology
Author(s)
Haller S., Falkovskiy P., Meuli R., Thiran J.P., Krueger G., Lovblad K.O., Kober T., Roche A., Marechal B.
ISSN
1432-1920 (Electronic)
ISSN-L
0028-3940
Publication state
Published
Issued date
11/2016
Peer-reviewed
Oui
Volume
58
Number
11
Pages
1153-1160
Language
english
Notes
Publication types: Journal Article
Publication Status: ppublish
Abstract
Automated brain MRI morphometry, including hippocampal volumetry for Alzheimer disease, is increasingly recognized as a biomarker. Consequently, a rapidly increasing number of software tools have become available. We tested whether modifications of simple MR protocol parameters typically used in clinical routine systematically bias automated brain MRI segmentation results.
The study was approved by the local ethical committee and included 20 consecutive patients (13 females, mean age 75.8 ± 13.8 years) undergoing clinical brain MRI at 1.5 T for workup of cognitive decline. We compared three 3D T1 magnetization prepared rapid gradient echo (MPRAGE) sequences with the following parameter settings: ADNI-2 1.2 mm iso-voxel, no image filtering, LOCAL- 1.0 mm iso-voxel no image filtering, LOCAL+ 1.0 mm iso-voxel with image edge enhancement. Brain segmentation was performed by two different and established analysis tools, FreeSurfer and MorphoBox, using standard parameters.
Spatial resolution (1.0 versus 1.2 mm iso-voxel) and modification in contrast resulted in relative estimated volume difference of up to 4.28 % (p < 0.001) in cortical gray matter and 4.16 % (p < 0.01) in hippocampus. Image data filtering resulted in estimated volume difference of up to 5.48 % (p < 0.05) in cortical gray matter.
A simple change of MR parameters, notably spatial resolution, contrast, and filtering, may systematically bias results of automated brain MRI morphometry of up to 4-5 %. This is in the same range as early disease-related brain volume alterations, for example, in Alzheimer disease. Automated brain segmentation software packages should therefore require strict MR parameter selection or include compensatory algorithms to avoid MR parameter-related bias of brain morphometry results.

Keywords
Aged, Algorithms, Artifacts, Brain/diagnostic imaging, Brain/pathology, Female, Humans, Image Enhancement/methods, Image Interpretation, Computer-Assisted/methods, Imaging, Three-Dimensional/methods, Magnetic Resonance Imaging/methods, Male, Organ Size, Pattern Recognition, Automated/methods, Reproducibility of Results, Sensitivity and Specificity, 3D T1, Hippocampus, MRI, Volumetry
Pubmed
Web of science
Open Access
Yes
Create date
23/09/2016 17:47
Last modification date
20/08/2019 16:21
Usage data