Premium
Reliability of fractional anisotropy in diffusion tensor imaging of healthy subjects in the ADNI database
Author(s) -
Chen Qianyun,
Abrigo Jill,
Chu Winnie C.W.
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.041494
Subject(s) - fractional anisotropy , diffusion mri , corpus callosum , white matter , superior longitudinal fasciculus , psychology , nuclear medicine , medicine , neuroscience , radiology , magnetic resonance imaging
Background Diffusion tensor imaging (DTI) detects white matter microstructural changes in early Alzheimer’s Disease (AD), and can potentially serve as a sensitive biomarker in longitudinal scans. In this study we determined the reliability of DTI‐derived fractional anisotropy (FA) in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset, which to our knowledge has not been reported. Method We selected 21 healthy subjects (12 women and 9 men; age range, 60‐89 years) from the ADNI2 with two sets of DTI performed within 3 months serving as surrogate test‐retest datasets. The subjects’ cognitive status remained stable within one year. The MRI was performed on the same 3.0 T GE scanner following ADNI standard protocol for DTI acquisition (2.7×2.7×2.7 mm 3 , b=1000 s/mm 2 , 5 b 0 and 41 diffusion weighted volumes). FMRIB Software Library (FSL) was used to process the DTI data to derive the FA scalar map. The two FA maps of the same subject were first co‐registered and then aligned to the JHU DTI atlas by using both linear and non‐linear registration method. A pre‐defined white matter region‐of‐interest (ROI) on the atlas was applied to the registered FA map to derive the value. The ROI related to AD were selected for analysis and included the corpus callosum (body, genu, splenium), corticospinal tract, superior longitudinal fasciculus, fronto‐occipital fasciculus and cingulum (cingulate gyrus, hippocampus). Reliability was assessed using intraclass correlation coefficient (ICC) and coefficient of variation (CoV) of the mean value within the ROI. Result Mean ICC of the paired FA values for all ROIs were 0.97(range: 0.93‐0.99). The CoVs for selected ROIs were all within 8% and the median was all below 2%. The greatest variation was found in the right cingulum hippocampus with CoV of 2.03%, while smallest variation 0.89% was present in the left superior longitudinal fasciculus. No significant correlation was found between ICC and CoV. Conclusion Using short interval longitudinal scans of healthy subjects in the ADNI dataset as surrogate test‐retest datasets, we found FA values of DTI exhibited good reliability. Our results lay the foundation for our future analysis of longitudinal DTI datasets in the ADNI database.