Please Wait...

DWI and DTI results on the normal controls, MCI and AD Subjects from the ROSAS study

L. Bracoud1, E. Bouguen2, F. Bonneville3, J. Schaerer1, V. Kiyasova2, F. Roche1, H. Basselerie3, H.M. Schneble2, J. Oh1, M. Pueyo2, J. Suhy1, B. Vellas3
1 BioClinica, Lyon, France and Newark, CA, USA, 2 Institut de Recherches Internationales Servier, Suresnes, France   3 CHU Toulouse, France

Diffusion-Weighted (DWI) and Diffusion-Tensor Imaging (DTI) assess microstructural brain tissue changes by measuring Apparent Diffusion Coefficient (ADC) and Fractional Anisotropy (FA) of water molecules.

As Alzheimer's disease (AD) progresses, neuronal degeneration and cerebral atrophy may manifest as an increase in ADC values owing to increased diffusivity of water molecules, and a decrease in FA values reflecting a loss in fiber tract integrity.

This work reports results of whole-brain histogram analysis applied to those data on Normal Controls (NC), MCI and AD subjects from the ROSAS study, a monocentric observational study  that was carried in Toulouse, France.


The ROSAS study is a monocentric observational study which ran in Toulouse, France, and was designed to identify and evaluate the clinical usefulness of AD biomarkers by collecting samples from Normal Controls (NC), Mild Cognitive Impairment (MCI) and AD subjects, following them for up to 4 years.

408 subjects aged 65 years or older were enrolled, including 110 Normal Controls (NC, no memory complaints, MMSE≥26 and CDR=0), 100 Mild Cognitive Impairment (MCI, MMSE≥24 and CDR=0.5, memory impairment based on RAVLT and who did not meet DSM IV criteria for AD dementia) and 198 AD (12≤MMSE≤26 and CDR≥0.5 and meeting DSM IV criteria).

Among MCI subjects, conversion to dementia during the course of the study was defined as an increase in CDR to >0.5.

MRI Data

DWI protocol used an echo-planar imaging in the axial plane on 31 slices of 4 mm with no inter-slice gap. The acquisition matrix was 112x89 (reconstructed at 256x256) over a FOV of 230x230 mm. 3 orthogonal directions were acquired, using 2 b values (0 and 1000 s/mm2).

DTI protocol used an echo-planar imaging in the axial plane on 50 slices of 2.2 mm with no inter-slice gap. The acquisition matrix was 100x100 (reconstructed at 256x256) over a FOV of 224x224 mm. 15 directions were acquired, using 2 b values (0 and 2500 s/mm2).

Scans were collected up to 3 times between Baseline and Month-48, at one site using a Philips Achieva 3T scanner, for consenting subjects. n=129, including 44 NC, 36 MCI and 49 AD. 14 MCI converted to AD during the study (MCI-c) while the others did not convert (MCI-nc).

Image processing

DWI ADC maps were used as generated on the scanner. DTI ADC and FA maps were centrally computed from the native DTI directional images.

All maps were quality controlled to discard scans exhibiting major artifacts (motion, ghosting, strong distortion, etc.).

Nonbrain voxels were automatically masked out using histogram-based thresholding and morphological operations performed on b0 images.

A CSF mask was determined using an upper cutoff applied to the ADC values at 16x10-4 mm2/s to account for and eliminate the increased water signal resulting from cerebral atrophy.

Data were also processed with no cutoff to keep the effect of cerebral atrophy.

Normalized histograms were generated for each visit. Histograms were characterized by mean, peak, standard deviation (SD) and height. Peak position was assessed by gamma-fitting of the histogram uptake to decrease sensitivity to noise.

Statistical analysis

Pearson's correlation between criteria was assessed at Baseline.

Differences between the 4 groups at Baseline were tested by a parametric approach (ANCOVA) adjusting for covariates (ApoE4, sex, age, education) and a non-parametric approach (Wilcoxon test) without adjustment for the covariates, as a robustness analysis. The pairwise differences were assessed based on the parametric model.

Global and pairwise differences between groups in the rate of changes across scans were tested using a parametric approach (Random Effects Model), adjusting for the same covariates.

No normalizing transformation of criteria were applied in parametric approaches.

For all comparisons between groups, a correction for multiplicity by FDR was applied.

Baseline correlations (see Fig. 1)

Comparing ADC results between DWI and DTI data at Baseline, mean and SD were highly correlated, while correlation on height was moderate, and weak on peak.

No strong correlation was found between FA and ADC parameters at Baseline.

Peak, mean and height were strongly correlated with each other at Baseline on DTI-ADC and DTI‑FA data. This was not as strong for DWI-ADC.

Differences in Baseline results among groups (see Fig. 2)

Baseline DTI-ADC mean, height and SD showed significant differences among groups. So did DWI-ADC mean and SD. This was driven by the significant pair-wise differences between AD and NC. Interestingly, DTI-ADC SD was able to distinguish NC from all groups.

No DTI-FA parameters showed significant differences.

Differences in changes across visits among groups (see Fig. 3)

Longitudinal changes in DTI-ADC mean, height and SD (without CSF) showed significant differences among groups. Mean DWI-ADC with CSF as well.

Similar to Baseline, those differences were mostly driven by the significant pair-wise differences between AD and NC. But interestingly again, DTI-ADC SD was able to distinguish NC from MCI-c and AD, and MCI-nc from MCI-c and AD.

Other DWI-ADC parameters and no DTI-FA parameters showed any significant changes between groups.

DTI-ADC standard deviation showed good  ability to differentiate NC from other groups at Baseline. Mean and height were only able to differentiate NC from  AD. Peak was not sensitive enough. DWI-ADC and DTI-FA showed limited value with such whole brain histogram approach.

DTI-ADC standard deviation also performed well longitudinally, in particular in distinguishing MCI non-converters from converters, which could be used to monitor treatment efficacy in a disease-modifying therapy.

While cross-sectional results may suffer from extending assessments into a multi-center setting, longitudinal results may be more robust to that respect.


Leader in Clinical Trial
Management Solutions

Successful clinical trials require the ability to see key details and uncover hidden insights. Bioclinica utilizes science and technology to bring clarity to clinical trials, helping companies to develop new life-improving therapies more efficiently and safely.

Integration of #eTMF and #CTMS eliminates redundancies, improves workflows & enables greater study control. Join th…
Bioclinica (Yesterday)
Thinking of building a study in-house? Here are the 5 things you should think about before starting a DIY study bui…
Bioclinica (6 days ago)
Integration of #eTMF and #CTMS eliminates redundancies, improves workflows & enables greater study control. Join th…
Bioclinica (1 week ago)
Your #IRT should not be the limiting factor during study start-up. The core features of Bioclinica's IRT agile plat…
Bioclinica (2 weeks ago)
Presenting Bioclinica CTMS: a rapid startup #CTMS that takes 2 weeks to implement, rather than months. Bioclinica C…
Bioclinica (2 weeks ago)
Improve doc & study mgmt. across all studies by integrating #eTMF & #CTMS. In this webinar, Bioclinica &…
Bioclinica (3 weeks ago)