Hippocampal subfields segmentation derived from Freesurfer 6.0: a multisite 3T reproducibility study in healthy elderly
Moira Marizzoni1, Daniele Orlandi1, Luigi Antelmi2, Flavio Nobili3, Mira Didic4,5, David Bartrés-Faz6, Ute Fiedler7, Peter Schonknecht8, Pierre Payoux9,10, Andrea Soricelli11,12, Alberto Beltramello13, Lucilla Parnetti14, Magda Tsolaki15, Paolo Maria Rossini16,17, Pieter Jelle Visser18, Regis Bordet19, Oliver Blin20, Giovanni Battista Frisoni1,21, Jorge Jovicich22, and on behalf of the PharmaCog Consortium1

1LENITEM Laboratory of Epidemiology, Neuroimaging, & Telemedicine — IRCCS San Giovanni di Dio-FBF, Brescia, Italy, 2Health Department, Foundation IRCCS Neurological Institute Carlo Besta, Milan, Italy, 3Department of Neuroscience, Ophthalmology, Genetics and Mother–Child Health (DINOGMI), University of Genoa, Genoa, Italy, 4APHM, CHU Timone, Service de Neurologie et Neuropsychologie, Marseille, France, 5Aix-Marseille Université, INSERM U 1106, Marseille, France, 6Department of Psychiatry and Clinical Psychobiology, Universitat de Barcelona and IDIBAPS, Barcelona, Spain, 7LVR-Clinic for Psychiatry and Psychotherapy, Institutes and Clinics of the University Duisburg-Essen, Essen, Germany, 8Department of Neuroradiology, University Hospital Leipzig, Leipzig, Germany, 9INSERM, Imagerie cérébrale et handicaps neurologiques, UMR 825, Toulouse, France, 10Université de Toulouse, UPS, Imagerie cérébrale et handicaps neurologiques, UMR 825, CHU Purpan, Place du Dr Baylac, Toulouse, France, 11IRCCS SDN, Naples, Italy, 12University of Naples Parthenope, Naples, Italy, 13Department of Neuroradiology, General Hospital, Verona, Italy, 14Section of Neurology, Centre for Memory Disturbances, University of Perugia, Perugia, Italy, 153rd Department of Neurology, Aristotle University of Thessaloniki, Thessaloniki, Greece, 16Dept. Geriatrics, Neuroscience & Orthopaedics, Catholic University, Policlinic Gemelli, Rome, Italy, 17IRCSS S.Raffaele Pisana, Rome, Italy, 18Department of Neurology, Alzheimer Centre, VU Medical Centre, Amsterdam, Italy, 19Department of Pharmacology, EA1046, University of Lille Nord de France, Lille, Italy, 20Pharmacology, Assistance Publique-Hôpitaux de Marseille, Aix-Marseille University-CNRS UMR 7289, Marseille, France, 21Memory Clinic and LANVIE - Laboratory of Neuroimaging of Aging, University Hospitals and University of Geneva, Geneva, Switzerland, 22Center for Mind/Brain Sciences, University of Trento, Rovereto, Italy

Synopsis

In this study we quantify the across-session reproducibility of hippocampus subfields obtained from the recently proposed ex-vivo atlas tool available in Freesurfer version 6.0. We use structural 3T multisite data from 65 healthy elderly participants scanned twice at least a week apart. We show that several subfields like Cornu Ammonis (CA) 1, hippocampal tail, molecular layer and subiculum offer, despite being smaller, comparable reliability errors to the whole hippocampus volume (2%). This suggests that these subfields may be valid and more specific markers to test disease progression in longitudinal studies, like for example Alzheimer's disease.

Purpose

With the new developments in neuroimaging techniques, there has been an increased interest in the consideration of hippocampus subfields studies using MRI data. The subfields package developed in Freesurfer 5.3 is one of the most used 1. Although its subfields segmentation was highly reproducible 2, its accuracy was recently questioned 3-5. A new and more accurate subfields package has been recently implemented in the last version of Freesurfer (v6.0) 6, However, the test-retest reproducibility across session of this tool has not yet been tested. The goal of this study was to extend our previous 40 subject study across 8 sites2 to evaluate the across-session reproducibility of the new Freesurfer 6.0 subfields segmentations relative to the whole hippocampus in 65 healthy elderly subjects across 13 sites.

Methods

Five healthy local volunteers (55-90 ys) were enrolled in 13 3T MRI sites (Siemens, GE, Philips) across Europe (Table 1) and were scanned in two sessions at least a week apart using the MRI acquisition protocol previously described 2. The morphometry analysis was performed on the average of two within-session T1 3D acquisitions using the longitudinal pipeline of Freesurfer v5.3 on the test and retest acquisitions. The subfield volumes were extracted with the new subfields package implemented in Freesurfer 6.0 6. For each site and ROI, volumes reliability was computed evaluating test-retest absolute differences relative to the mean (absolute error).

Results

Subfields analysis was focused on: Cornu Ammonis (CA) 1, CA2-3, CA4, subiculum, presubiculum, parasubiculum, granule cell layer of dentate gyrus (GC-ML-DG), fimbria and hippocampal fissure, hippocampus-amygdala-transition-area (HATA), molecular layer and the hippocampal tail. The reproducibility error of the whole hippocampus (about 2%) was similar to that of several subfields, like CA1, hippocampal tail, molecular layer and subiculum. Other structures had higher reproducibility errors, about 4% for presubiculum, GC-ML-DG, CA4 and CA2-3 and around 8% for fimbria, hippocampal fissure, parasubiculum and HATA. Direct comparisons of reproducibility across Freesurfer versions are difficult because the subfield segmentations cover different hippocampus areas even if they have the same label names. Smaller volumes tend to have higher reproducibility errors.

Discussion and Conclusions

Despite the differences of the 13 MRI scanner configurations we found good and consistent hippocampal subfields reproducibility for CA1, CA2-3, CA4, hippocampal tail, molecular layer, subiculum, presubiculum and GC-ML-DG. Our findings suggest that the larger hippocampal subfields volume may be more accurate and reliable longitudinal markers in multisite studies, offering higher biological specificity to the characterization of disease progression.

Acknowledgements

Pharmacog is funded by the EU-FP7 for the Innovative Medicine Initiative (grant n°115009).

References

1 Van Leemput K, Bakkour A, Benner T, et al. Automated segmentation of hippocampal subfields from ultra-high resolution in vivo MRI. Hippocampus. 2009;19(6):549-557. 2 Marizzoni M, Antelmi L, Bosch B, et al. Longitudinal reproducibility of automatically segmented hippocampal subfields: A multisite European 3T study on healthy elderly. Hum Brain Mapp. 2015;36(9):3516-3527. 3 Wisse LE, Biessels GJ, Geerlings MI. A Critical Appraisal of the Hippocampal Subfield Segmentation Package in FreeSurfer. Front Aging Neurosci. 2014;25;6:261. 4 de Flores R, La Joie R, Chételat G. Structural imaging of hippocampal subfields in healthy aging and Alzheimer's disease. Neuroscience. 2015;19;309:29-50. 5 Yushkevich PA, Amaral RS, Augustinack JC, et al. Quantitative comparison of 21 protocols for labeling hippocampal subfields and parahippocampal subregions in in vivo MRI: towards a harmonized segmentation protocol. Neuroimage. 2015; 1;111:526-541.6 Iglesias JE, Augustinack JC, Nguyen K et al. A computational atlas of the hippocampal formation using ex vivo, ultra-high resolution MRI: Application to adaptive segmentation of in vivo MRI. Neuroimage. 2015;15;115:117-137.

Figures

Table 1: Demographic and MRI system differences across MRI sites.

FIGURE 1: Whole hippocampus and its subfields volume reproducibility for the 13 MRI sites (3T) averaging the within session T1 images. Within-site group mean and standard deviation of reproducibility error averaged across subjects and hemispheres. The last black column shows for each region the averaged across sites. No significant MRI site effects were found (KW test, p> .05).



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1930