Please Wait...

Longitudinal Volumetric Changes in Controls, MCI and Ad Subjects from the ROSAS Study, as Compared to ADNI2

Luc Bracoud 1, Hans-Martin Schneble 2, Raluca Gramada 3, Fabrice Bonneville 3, Florent Roche 1, Isabelle Guignot 2, Joël Schaerer 1, Françoise Lala 3, Nathalie Sastre 3, Pierre-Jean Ousset 3, Maria Pueyo 2, Chahin Pachai 1, Bruno Vellas 3 and the Alzheimer's Disease Neuroimaging Initiative 1 BioClinica, Lyon, France   2 Institut de Recherches Internationales Servier, Suresnes, France   3 C.H.U. Toulouse, France - email:


The ROSAS study is a monocentric observational study running in Toulouse, France, 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 over up to 4 years.

In this work, changes over time in whole brain, lateral ventricles and hippocampal volumes were assessed for the ROSAS cohort. For whole brain and lateral ventricles, 2 independent methods (Boundary Shift Integral - BSI and Tensor Based Morphometry - TBM) were applied, for comparison purposes.

These atrophy rates were also compared to those from the ADNI2 cohort, as the imaging context of ADNI2 is identical to ROSAS.

In addition, we looked at relationship between initial hippocampal volume and longitudinal change (atrophy rate and conversion to AD).


ROSAS data
Subjects aged 65 years or older were enrolled in the study, including NC (no objective memory impairments, MMSE≥26 and CDR=0), MCI (MMSE≥24 and CDR=0.5, memory impairment based on RAVLT and who did not meet DSM-IV-TR criteria for AD dementia and AD (12≤MMSE≤26 and CDR≥0.5 and meeting DSM-IV-TR criteria).

Subjects with other types of dementia and/or cognitive issues not related to AD were excluded.
3D T1-weighted (3DT1) MRI scans were collected at Baseline and up to twice between Month 12 and Month 48, at one site using a Philips Achieva 3T scanner, for consenting subjects.

3DT1 data consisted of a sagittal 3DTFE sequence with 1 mm thick slices and a 175x256 acquisition matrix over a square FOV of 256 mm.

Only subjects with good quality scans (no significant imaging artifacts, same MRI protocol used across timepoints) were considered for analysis.

For subjects who underwent a change in the MRI protocol between their first and second scans, atrophy was only assessed between the second and third scans.

NC subjects who converted to MCI or AD during the course of the study were discarded.
MCI subjects were further classified as non-converters or converters (MCI-nc/MCI-c).

A total of 101 subjects were considered, including 35 NC, 15 MCI-nc, 13 MCI-c and 38 AD.

ADNI2 data
Baseline, Months 3, 6, 12 and 24 data from 593 subjects from the ADNI2 cohort were also assessed for comparison purposes.

All subjects had been scanned on 3T scanners, providing a comparable imaging context with ROSAS data.

Those subjects were separated into the following categories: Normal Controls (NC, n=163), Early/Late Mild Cognitive Impairment (EMCI/LMCI, n=161/156) and AD (n=113), based on ADNI criteria.

NC who converted to MCI or AD within 36 months were discarded.

Image processing The following fully-automated MRI measures were performed on all data:
The volumes of whole brain and lateral ventricles were assessed at Baseline using an atlas segmentation algorithm [1].

Follow-up scans were automatically spatially realigned against Baseline scans using rigid and affine registration algorithms. Registration was performed in a so-called midway space, in order to prevent bias towards either scan (symmetric process) [2].

Volume changes at follow-up timepoints were assessed using Boundary Shift Integral (BSI) [3] and Tensor-Based Morphometry (TBM) [4].

Hippocampal volume (HCV) was assessed at Baseline using a multi-atlas segmentation algorithm [5].

Volume changes at follow-up timepoints were assessed using Hippocampal Boundary Shift Integral [6].

Statistical Analysis
Annualized volume loss was estimated across all available timepoints by fitting a linear model by robust regression using an M estimator. No adjustment against clinical characteristics was performed.

Agreement between techniques were evaluated using Intraclass Correlation (ICC) coefficient ρ.


Annualized mean volume changes and standard errors are summarized in the below tables. Table 1 shows results on the ROSAS cohort using BSI and TBM methods, for brain (BBSI/BTBM), ventricles (VBSI/VTBM) and hippocampal (HBSI) volume changes. Table 2 shows similar results on the ADNI2 cohort using BSI.

A strong agreement was shown between BSI and TBM (ρ ≥ 0.97 with lower bound of 95%CI  ≥ 0.96, see Fig. 1) and led to the same volume changes in whole brain and lateral ventricles in ROSAS NC, MCI-nc, MCI-c and AD.

When assessing volume changes in whole brain, lateral ventricles and hippocampus (see Fig. 2), ROSAS controls showed less atrophy than ADNI2 controls, which is likely explained by stricter inclusion criteria (MMSE ≥ 26 for ROSAS and ≥ 24 for ADNI2). In addition, MCI subgroups always followed the same pattern of atrophy rate for all structures: EMCI (ADNI2) < MCI-nc (ROSAS) < LMCI (ADNI2) < MCI-c (ROSAS), while ROSAS and ADNI2 AD were similar.

Baseline HCV was weakly correlated to brain atrophy and moderately correlated to hippocampal atrophy (see Fig. 3).

When looking at MCI subjects in particular, a cut-off HCV of 7000 mm3 was significantly associated with conversion to AD (Table 3).

The ROSAS cohort provides additional reference values for the design of future clinical trials.


Belaroussi et al., Labeling of brain MRI images using atlas propagation and classification-based nearest-neighbor transform, Poster session, AAN 2011

Leung et al., Consistent multi-time-point brain atrophy estimation from the boundary shift integral, Neuroimage 2012

Freeborough et al., The boundary shift integral: an accurate and robust measure of cerebral volume changes from registered repeat MRI, IEEE Trans Med Imaging, 1997

Vercauteren et al., Symmetric log-domain diffeomorphic Registration: a demons-based approach, Med Image Comput Comput Assist Interv, 2008

Belaroussi et al., Multi-Atlas hippocampus segmentation refined with intensity-based tissue classification, Poster session, AAIC 2012

Barnes et al., Automatic calculation of hippocampal atrophy rates using a hippocampal template and the boundary shift integral, Neurobiology of aging, 2006


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.

Still time to catch up on summer reading on recent developments in #Alzheimer's Disease Fluid Biomarkers here int t…
bioclinica (3 days ago)
Our clients have substantial success w #eSource, some spanning well over a decade. Still other organizations remain…
bioclinica (4 days ago)
Our clients have substantial success w #eSource, some spanning well over a decade. Still some in the industry remai…
bioclinica (6 days ago)
RT @RfwrightLSL: When Do You Think #BYOD (Bring Your Own Device) Will Become Widely Used To Capture PRO Data In #Clinical #Research? https:…
bioclinica (2 weeks ago)
ICYMI: Our V.P./Head of Neuroscience & Cardiovascular Svcs., Dr. Joyce Suhy, talks about challenges in determining…
bioclinica (2 weeks ago)
RT @FCR_News_Today: FDA Issues New Guides on Use of Electronic Health Information in #ClinicalTrials. #ClinicalResearch…
bioclinica (2 weeks ago)

Latest Blogs:

Latin America: Benefit from the Right Partner
Removing Risk from Clinical Trial Management System (CTMS) Implementations
Collaboration Between Clinical Operations and the Logistics and Supply Chain Teams is Key to Trial Success
The Value of Protocol Review
CTMS and RBM: Hot Topics at OCT Nordics in Copenhagen