Free water elimination using a bi-tensor model improves test-retest reproducibility of diffusion tensor imaging indices in the brain: a longitudinal multisite reliability study of healthy elderly subjects

Angela Albi¹, Ofer Pasternak², Ludovico Minati^1,3, Moira Marizzoni⁴, Giovanni Frisoni^4,5, David Bartrés-Faz⁶, Núria Bargalló⁷, Beatriz Bosch⁸, Paolo Maria Rossini^9,10, Camillo Marra¹¹, Bernhard Müller¹², Ute Fiedler¹², Jens Wiltfang^12,13, Luca Roccatagliata^14,15, Agnese Picco¹⁶, Flavio Mariano Nobili¹⁶, Oliver Blin¹⁷, Julien Sein¹⁸, Jean-Philippe Ranjeva¹⁸, Mira Didic^19,20, Stephanie Bombois²¹, Renaud Lopes²¹, Régis Bordet²¹, Hélène Gros-Dagnac^22,23, Pierre Payoux^22,23, Giada Zoccatelli²⁴, Franco Alessandrini²⁴, Alberto Beltramello²⁴, Antonio Ferretti^25,26, Massimo Caulo^25,26, Marco Aiello²⁷, Carlo Cavaliere²⁷, Andrea Soricelli^27,28, Lucilla Parnetti²⁹, Roberto Tarducci³⁰, Piero Floridi³¹, Magda Tsolaki³², Manos Constantinidis³³, Antonios Drevelegas³⁴, and Jorge Jovicich¹

¹Center for Mind/Brain Sciences (CIMEC), University of Trento, Rovereto, Trento, Rovereto (Trento), Italy, ²Departments of Psychiatry and Radiology, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts, Boston, MA, United States, ³Scientific Department, Fondazione IRCCS Istituto Neurologico Carlo Besta, Milan, Italy, Milan, Italy, ⁴LENITEM Laboratory of Epidemiology, Neuroimaging, & Telemedicine — IRCCS San Giovanni di Dio-FBF, Brescia, Italy, Brescia, Italy, ⁵Memory Clinic and LANVIE, Laboratory of Neuroimaging of Aging, University Hospitals and University of Geneva, Geneva, Switzerland, Geneva, Switzerland, ⁶Department of Psychiatry and Clinical Psychobiology, Universitat de Barcelona and IDIBAPS, Barcelona, Spain, Barcelona, Spain, ⁷Department of Neuroradiology and Magnetic Resonance Image core Facility, Hospital Clínic de Barcelona, IDIBAPS, Barcelona, Spain, Barcelona, Spain, ⁸Alzheimer's Disease and Other Cognitive Disorders Unit, Department of Neurology, Hospital Clínic, and IDIBAPS, Barcelona, Spain, Barcelona, Spain, ⁹Deptartment Geriatrics, Neuroscience & Orthopaedics, Catholic University, Policlinic Gemelli, Rome, Italy, Rome, Italy, ¹⁰IRCSS S.Raffaele Pisana, Rome, Italy, Rome, Italy, ¹¹Center for Neuropsychological Research, Catholic University, Rome, Italy, Rome, Italy, ¹²LVR-Clinic for Psychiatry and Psychotherapy, Institutes and Clinics of the University Duisburg-Essen, Essen, Germany, Essen, Germany, ¹³Department of Psychiatry and Psychotherapy, University Medical Center (UMG), Georg August University, Göttingen, Germany, Göttingen, Germany, ¹⁴Department of Neuroradiology, IRCSS San Martino University Hospital and IST, Genoa, Italy, Genoa, Italy, ¹⁵Department of Health Sciences, University of Genoa, Genoa, Italy, Genoa, Italy, ¹⁶Department of Neuroscience, Ophthalmology, Genetics and Mother–Child Health (DINOGMI), University of Genoa, Genoa, Italy, Genoa, Italy, ¹⁷Pharmacology, Assistance Publique — Hôpitaux de Marseille, Aix-Marseille University — CNRS, UMR 7289, Marseille, France, Marseille, France, ¹⁸CRMBM–CEMEREM, UMR 7339, Aix Marseille Université — CNRS, Marseille, France, Marseille, France, ¹⁹APHM, CHU Timone, Service de Neurologie et Neuropsychologie, Marseille, France, Marseille, France, ²⁰Aix Marseille Université, Inserm, INS UMR_S 1106, 13005, Marseille, France, Marseille, Italy, ²¹Université de Lille, Inserm, CHU Lille, U1171 - Degenerative and vascular cognitive disorders, F-59000 Lille, France, Lille, France, ²²INSERM, Imagerie cérébrale et handicaps neurologiques, UMR 825, Toulouse, France, Toulouse, France, ²³Université de Toulouse, UPS, Imagerie cérébrale et handicaps neurologiques, UMR 825, CHU Purpan, Place du Dr Baylac, Toulouse Cedex 9, France, Toulouse, France, ²⁴Department of Neuroradiology, General Hospital, Verona, Italy, Verona, Italy, ²⁵Department of Neuroscience Imaging and Clinical Sciences, University “G. d'Annunzio” of Chieti, Italy, Chieti, Italy, ²⁶Institute for Advanced Biomedical Technologies (ITAB), University “G. d'Annunzio” of Chieti, Italy, Chieti, Italy, ²⁷IRCCS SDN, Naples, Italy, Naples, Italy, ²⁸University of Naples Parthenope, Naples, Italy, Naples, Italy, ²⁹Section of Neurology, Centre for Memory Disturbances, University of Perugia, Perugia, Italy, Perugia, Italy, ³⁰Medical Physics Unit, Perugia General Hospital, Perugia, Italy, Perugia, Italy, ³¹Neuroradiology Unit, Perugia General Hospital, Perugia, Italy, Perugia, Italy, ³²3rd Department of Neurology, Aristotle University of Thessaloniki, Thessaloniki, Greece, Thessaloniki, Greece, ³³Interbalkan Medical Center of Thessaloniki, Thessaloniki, Greece, Thessaloniki, Greece, ³⁴Interbalkan Medical Center of Thessaloniki, Thessaloniki, Greece Department of Radiology, Aristotle University of Thessaloniki, Thessaloniki, Greece, Thessaloniki, Greece

Synopsis

Brain diffusion tensor imaging (DTI) provides in-vivo characterization of white matter tissue microstructure. In this study we demonstrate that free water elimination in brain diffusion MRI significantly improves the test-retest reproducibility of DTI metrics (fractional anisotropy, axial, radial and mean diffusivity) in a multsite 3T setting. This work has important clinical applications since the improved reliability may provide increased sensitivity in longitudinal studies quantifying white matter neurophysiological processes related to disease stage/progression and treatment responses.

Introduction

Diffusion MRI is a non-invasive tool able to provide unique in-vivo microstructural brain information, revealing white matter structure and organization.¹ Common clinical measures characterizing white matter are inferred from diffusion tensor imaging (DTI), such as Fractional Anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD) and axial diffusivity (AD). Longitudinal assessment of these metrics is of interest as potential biomarkers of disease prediction/progression and treatment response, particularly in large multisite studies.² The sensitivity of longitudinal studies is typically limited by between session test-retest reproducibility. One factor that may influence reliability is partial volume of white matter tissue with a fast diffusion component of water molecules that are free to diffuse, such as cerebrospinal fluid (CSF) next to the ventricles and around the brain parenchyma.³ A two-tensor model for free water elimination (FWE) has been recently proposed to correct for this partial volume. The FWE method provides corrected DTI measures, which are expected to be more tissue-specific than the non-corrected DTI measures. However, the reproducibility of FWE comparing with DTI has not yet been tested. Here we evaluate the test-retest longitudinal reproducibility of FWE measures by extending a previous multi-site reliability study⁴ that evaluated DTI measures in healthy elderly subjects. We calculate the test-retest reproducibility using the FWE method and compare the results with and without FWE for FA, MD, AD and RD on 48 white matter areas.

Methods

Ten clinical European 3T MRI sites using Philips, GE and Siemens scanners participated in this study, each site had a single MRI scanner. Each site recruited 5 volunteers in the age range of 50-80 years with no history of major psychiatric, neurological or cognitive impairment. Participants were scanned in two sessions (test and retest) between 7 days and a maximum of 60 days apart. A calibrated acquisition protocol based on vendor sequences was used (2x2x2 mm³ voxels, axial slices with no gaps, fat suppression, 5 b0 volumes, b=700 s/mm², 30 DWI).⁴ Two subjects were excluded for missing data. Data quality control and preprocessing procedures were as described in the original study⁴ and followed by calculating the scalar maps (FA, AD, RD, MD) from both the standard DTI and from the bi-tensor FWE model.³ Separate white matter skeletons for each site were generated with standard TBSS analyses⁵, and all the scalar maps were projected to the skeleton. To estimate the reproducibility error of specific white matter areas, the different measures were averaged over the pre-defined ROIs. We used the JHU-ICBM-FA-1mm atlas overlapped with the corresponding site's FA skeleton. No white matter lesions were observed, therefore we combined metrics over the right and left hemisphere for each subject, resulting with 27 ROIs. All scalar maps were averaged within the ROIs. To test for test-retest longitudinal reproducibility we used a one-way Kruskall-Wallis test. To test for difference in reproducibility between uncorrected and FWE-corrected maps we used the two-tailed Wilcoxon sign-rank test, corrected for multiple comparisons with False Discovery Rate (FDR).^6,7

Results

Averaging over all ROIs, FWE increased FA values by 0.11±0.01, lowered MD values by 0.22±0.02, lowered RD values by 0.21±0.02 and lowered AD values by 0.22±0.02. Figure 1 shows color-coded difference maps (Figure 1c). The test-retest reproducibility errors for all metrics showed no significant sites effects (Kruskall-Wallis, p > 0.05), we therefore grouped the reproducibility metrics across sites with and without the FWE correction (e.g. see FA in Table 1). The reproducibility errors of the two methods were compared using a two-tailed Wilcoxon sign-rank test by grouping scores from all 48 subjects (Table 2). Test-retest errors were significantly reduced in most white matter ROIs (Wilcoxon test, p < 0.05 with FDR). From the 27 ROIs, reproducibility errors were reduced in 15 ROIs for AD (56%), 14 ROIs for RD (52%), 19 ROIs for MD (70%) and 20 ROIs for FA (74%). In six ROIs FWE correction resulted with higher reproducibility error (Table 2).

Discussion and Conclusions

FWE produced changes in diffusion scalar metrics that were consistent with previous studies.^3,7 FWE also showed significant test-retest reliability improvements of diffusion scalar estimates relative to the standard single tensor estimates in most brain areas tested. The improved reliability of FWE is consistent with a reduced contribution of partial volume effects, which may arise due to variability in both subject positioning and extracellular components (e.g. dehydration, inflammation) across scanning sessions. In this study we showed how the removal of the fast diffusing component can increase sensitivity to longitudinal changes in white matter characteristics by reducing reproducibility errors, thus supporting the choice of FWE as a useful and important additional step for DTI data processing.

Acknowledgements

Pharmacog is funded by the EU-FP7 for the Innovative Medicine Initiative (grant n°115009).
This work was partially supported by the following NIH grants: R01MH074794, 2P41EB015902 , 1R01AG042512, R01MH102377, and by a NARSAD young investigator award.

References

1. Basser, P. J., & Pierpaoli, C. (1996). Microstructural and physiological features of tissues elucidated by quantitative-diffusion-tensor MRI. Journal of Magnetic Resonance, 111, 209-219.

2. Horn, J. Van, & Toga, A. (2009). Multi-site neuroimaging trials. Current Opinion in Neurology, 22(4), 370–378.

3. Pasternak, O., Sochen, N., Gur, Y., Intrator, N., & Assaf, Y. (2009). Free water elimination and mapping from diffusion MRI. Magnetic Resonance in Medicine, 62, 717–730.

4. Jovicich, J., Marizzoni, M., Bosch, B., Bartrés-Faz, D., Arnold, J., Benninghoff, J., … Frisoni, G. B. (2014). Multisite longitudinal reliability of tract-based spatial statistics in diffusion tensor imaging of healthy elderly subjects. NeuroImage, 101, 390–403.

5. Smith, S. M., Jenkinson, M., Johansen-Berg, H., Rueckert, D., Nichols, T. E., Mackay, C. E., … Behrens, T. E. J. (2006). Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. NeuroImage, 31, 1487–505.

6. Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society.

7. Benjamini, Y., & Yekutieli, D. (2001). The control of the false discovery rate in multiple testing under depencency. The Annals of Statistics, 29(4), 1165–1188.

8. Berlot, R., Metzler-Baddeley, C., Jones, D. K., & O’Sullivan, M. J. (2014). CSF contamination contributes to apparent microstructural alterations in mild cognitive impairment. NeuroImage, 92, 27–35.

Figures

Table 1. Test-retest reproducibility error averaged across subjects (mean±SD) for non-corrected and FWE-corrected FA across the 27 white matter labels of the JHU-ICBM atlas.

Table 2. FWE effect on DTI measures reproducibility. Each cell shows FDR corrected p-values of test-retest reproducibility error difference between FWE and non-corrected DTI metrics. White cells denote no significant differences, grey cells denote significant improvement in reproducibility with FWE, and black cells denote significant worsening of reproducibility with FWE.

Figure 1. Example data showing FA, MD, RD and AD for (a) uncorrected volumes and (b) FWE-corrected volumes. The differences are highlighted in absolute difference maps (c) (uncorrected map subtracted from FWE-corrected map)

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)

1043