Premium
Brain imaging data quality and denoising: Impact on the study of sex differences in Alzheimer’s disease
Author(s) -
Caldwell Jessica ZK,
Yang Zhengshi,
Kaplan Nikki,
Leavitt MacKenzie,
Miller Justin,
Mishra Virendra R,
Shan Guogen,
Cordes Dietmar
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.046684
Subject(s) - neuroimaging , alzheimer's disease neuroimaging initiative , medicine , magnetic resonance imaging , resting state fmri , spatial normalization , white matter , disease , psychology , dementia , audiology , psychiatry , pathology , radiology
Abstract Background Recent research highlights sex effects in Alzheimer’s disease (AD). However, examining sex effects presents challenges, for example due to Magnetic Resonance Imaging (MRI) methodological differences and effects of sex on resting state functional MRI (rsfMRI) being small compared to effects of factors like age. Attention to data processing methodology may help to address this challenge. The present discussion will highlight the results of our recent examinations of the effect of sex on success of automated MRI volumetric analysis with clinically‐obtained MRI, and the role of rsfMRI data denoising on measuring sex effects in AD. Method Volumetric quality was assessed in 438 patients seen in an outpatient neurology clinic specializing in diagnosis and treatment for neurodegenerative diseases (47.7% women), and rsfMRI in 168 participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI: 49 Cognitively Normal (CN); 67 early Mild Cognitive Impairment (eMCI); 52 AD; 52% women). MRI data were processed using Freesurfer 6.0, with in‐house quality assessment. Resting state data were preprocessed with realignment, slice‐time correction, coregistration and normalization to standard MNI space, and then denoised using traditional nuisance regression methods with global signal, mean white matter and cerebrospinal fluid time series, and motion parameters; Compcor; and a novel in‐house developed, deep neural network approach. Result In‐house quality assessment resulted in discarding 43% of segmented images. Individuals whose images were discarded were significantly older than individuals with retained images, but sex did not impact retention. In contrast, using our novel rsfMRI denoising technique, CN women showed significantly lower degree centrality, global efficiency, local efficiency, and clustering coefficient, and significantly longer path length. Differences were less significant in MCI and not significant in AD. Traditional denoising approaches revealed weak differences or no differences by sex. Conclusion These findings suggest that quality control approaches to clinical imaging processing may be important to the study of AD generally, but that data denoising techniques in rsfMRI may have impact specifically on the study of AD‐related sex differences.