Event Abstract

Back to Event

A hippocampal volume-based biomarker for use in multi-cohort, heterogeneous MRI data sets.

Rembrandt Bakker1, 2*, Daniel Fängström1, Marcel Koek3, Elena Ranguelova4, Olga Meulenbroek5, Saartje Burgmans6, Hugo Vrenken7, Wiro Niessen3 and Paul H. Tiesinga1
  • 1 Radboud University Nijmegen, Dept. of Neuroinformatics, Donders Institute for Brain, Cognition and Behaviour, Netherlands
  • 2 Research Center Jülich, Institute of Neuroscience and Medicine INM-6/IAS-6, Germany
  • 3 Erasmus MC, Biomedical Imaging Group Rotterdam, Depts. Medical Informatics and Radiology, Netherlands
  • 4 Netherlands eScience Center, Netherlands
  • 5 Radboudumc, Dept. of Geriatric Medicine/Radboud Alzheimer Centre, Donders Institute for Brain, Cognition and Behaviour, Netherlands
  • 6 Maastricht University, Dept. of Psychiatry and Neuropsychology/Alzheimer Center Limburg, Netherlands
  • 7 VU University Medical Center, Dept. of Radiology and Nuclear Medicine, Neuroscience Campus Amsterdam, Netherlands

Data sharing of neuroimaging cohort studies has the potential benefit to boost statistical power when it comes to detecting correlations between image-derived biomarkers and (future) diagnose. However, while data within cohorts is obtained with MRI protocols that are fixed for the duration of the study, MRI protocols vary widely between cohorts. The same holds true for the manual segmentations that are produced to drive the multi-atlas registration part of the analysis, as shown in Fig. 1 which compares the average manually segmented hippocampus for each cohort (Ams - Amsterdam, Maa - Maastricht, Nij - Nijmegen, Rot - Rotterdam; Red/Green denotes missing/extra part compared to overall average) to the overall average. This increases bias and variance of the image-derived biomarkers, and partially offsets the benefits of data sharing. This study focuses on modifying an existing pipeline (Van der Lijn et al. 2008, Van der Lijn et al. 2012) for hippocampal volume prediction in such a way that it becomes much less dependent of the MRI protocol. The original pipeline uses multi-atlas registration to construct a probabilistic atlas, followed by a graphcuts-based segmentation that optimizes an energy function that contains this probabilistic atlas, as well as image-intensity and smoothness components. It achieves a DICE score of 0.86 on the test set of 20 manual segmentations. In our new approach, we improve the robustness of the multi-atlas registration by using a brainmask, histogram equalization within this mask, and image blanking outside the masked region. We derive a measure to compute volumes and DICE scores directly from the probabilistic atlas, i.e. without computing the actual shape of the hippocampus. Without applying graphcuts, the modified pipeline achieves a DICE score of 0.85, i.e. slightly lower than the full pipeline but with far fewer parameters that need to be tuned. The slope of the fitted line between predicted and manually segmented hippocampal volume is consistent across the four different cohorts, with a mean value of 0.46, well above the previously reported value of 0.37 (Van der Lijn et al. 2008, multi-atlas method without graphcuts). The pipeline uses the eScience infrastructure described in De Boer et al. (2013), and its parallel execution is governed by a newly developed, python based pipeline engine (https://github.com/rbakker/FancyPipe).

Figure 1
Image

Acknowledgements

Supported by the Netherlands eScience Center, grant 027.011.304

References

1. van der Lijn F, den Heijer T, Breteler MM, Niessen WJ (2008) Hippocampus segmentation in MR images using atlas registration, voxel classification, and graph cuts. Neuroimage 43, 708-20. doi: 10.1016/j.neuroimage.2008.07.058.

2. van der Lijn F, de Bruijne M, Klein S, den Heijer T, Hoogendam YY, van der Lugt A, Breteler MM, Niessen WJ (2012) Automated brain structure segmentation based on atlas registration and appearance models. IEEE Trans Med Imaging 31(2):276-86. doi: 10.1109/TMI.2011.2168420.

3. De Boer P, Ranguelova E, Ivanova M, Koek M, Van Der Lijn F, Niessen W, Versteeg A, Vrenken H, Burgmans S, Van Boxtel M, Meulenbroek O, De Leeuw F, Bakker R and Tiesinga P (2013). eScience Infrastructure for sharing neuroimaging data and running validated analysis pipelines on a high performance cloud. Front. Neuroinform. Conference Abstract: Neuroinformatics 2013. doi: 10.3389/conf.fninf.2013.09.00085.

Keywords: Neuroimaging methods, atlas-based segmentation, Multi-center, Hippocampus, Histogram equalization, graph cuts

Conference: Second Belgian Neuroinformatics Congress, Leuven, Belgium, 4 Dec - 4 Dec, 2015.

Presentation Type: Demonstration

Topic: Brain Imaging

Citation: Bakker R, Fängström D, Koek M, Ranguelova E, Meulenbroek O, Burgmans S, Vrenken H, Niessen W and Tiesinga PH (2015). A hippocampal volume-based biomarker for use in multi-cohort, heterogeneous MRI data sets.. Front. Neuroinform. Conference Abstract: Second Belgian Neuroinformatics Congress. doi: 10.3389/conf.fninf.2015.19.00042

Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.

The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.

Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.

For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.

Received: 14 Nov 2015; Published Online: 17 Nov 2015.

* Correspondence: Dr. Rembrandt Bakker, Radboud University Nijmegen, Dept. of Neuroinformatics, Donders Institute for Brain, Cognition and Behaviour, Nijmegen, Netherlands, r.bakker@donders.ru.nl