Benchmarking Reconstruction Methods for Bundle Segmentation: Preparing for Clinical Translation
Presented During: Poster Session 3
Friday, June 27, 2025: 01:45 PM - 03:45 PM
Friday, June 27, 2025: 01:45 PM - 03:45 PM
Presented During: Poster Session 4
Saturday, June 28, 2025: 01:45 PM - 03:45 PM
Saturday, June 28, 2025: 01:45 PM - 03:45 PM
Poster No:
1283
Submission Type:
Abstract Submission
Authors:
Amelie Rauland1,2, Steven Meisler3, Aaron Alexander-Bloch4,5,6, David Roalf4,5, Erica Baller3,5, Audrey Luo3,4,5, Joëlle Bagautdinova3,5, Kathrin Reetz7, Oleksandr Popovych1,8, Raquel Gur4,5, Ruben Gur4,5, Simon Eickhoff1,8, Matthew Cieslak3,5, Theodore Satterthwaite3,5
Institutions:
1Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich, Jülich, Germany, 2Department of Psychiatry, Psychotherapy and Psychosomatics, Faculty of Medicine, RWTH Aachen University, Aachen, Germany, 3The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania, Philadelphia, USA, 4Lifespan Brain Institute (LiBI) of Penn Medicine and CHOP, University of Pennsylvania, Philadelphia, USA, 5Department of Psychiatry, University of Pennsylvania, Philadelphia, USA, 6Department of Child and Adolescent Psychiatry and Behavioral Science, The Children’s Hospital of Philadelphia, Philadelphia, USA, 7Department of Neurology, RWTH Aachen University, Aachen, Germany, 8Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf, Heinrich Heine University Düsseldorf, Düsseldorf, Germany
First Author:
Amelie Rauland
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Department of Psychiatry, Psychotherapy and Psychosomatics, Faculty of Medicine, RWTH Aachen University
Jülich, Germany|Aachen, Germany
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Department of Psychiatry, Psychotherapy and Psychosomatics, Faculty of Medicine, RWTH Aachen University
Jülich, Germany|Aachen, Germany
Co-Author(s):
Steven Meisler, PhD
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania
Philadelphia, USA
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania
Philadelphia, USA
Aaron Alexander-Bloch
Lifespan Brain Institute (LiBI) of Penn Medicine and CHOP, University of Pennsylvania|Department of Psychiatry, University of Pennsylvania|Department of Child and Adolescent Psychiatry and Behavioral Science, The Children’s Hospital of Philadelphia
Philadelphia, USA|Philadelphia, USA|Philadelphia, USA
Lifespan Brain Institute (LiBI) of Penn Medicine and CHOP, University of Pennsylvania|Department of Psychiatry, University of Pennsylvania|Department of Child and Adolescent Psychiatry and Behavioral Science, The Children’s Hospital of Philadelphia
Philadelphia, USA|Philadelphia, USA|Philadelphia, USA
David Roalf
Lifespan Brain Institute (LiBI) of Penn Medicine and CHOP, University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
Lifespan Brain Institute (LiBI) of Penn Medicine and CHOP, University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
Erica Baller
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
Audrey Luo
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania|Lifespan Brain Institute (LiBI) of Penn Medicine and CHOP, University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA|Philadelphia, USA
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania|Lifespan Brain Institute (LiBI) of Penn Medicine and CHOP, University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA|Philadelphia, USA
Joëlle Bagautdinova
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
Oleksandr Popovych
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf, Heinrich Heine University Düsseldorf
Jülich, Germany|Düsseldorf, Germany
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf, Heinrich Heine University Düsseldorf
Jülich, Germany|Düsseldorf, Germany
Raquel Gur
Lifespan Brain Institute (LiBI) of Penn Medicine and CHOP, University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
Lifespan Brain Institute (LiBI) of Penn Medicine and CHOP, University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
Ruben Gur
Lifespan Brain Institute (LiBI) of Penn Medicine and CHOP, University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
Lifespan Brain Institute (LiBI) of Penn Medicine and CHOP, University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
Simon Eickhoff
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf, Heinrich Heine University Düsseldorf
Jülich, Germany|Düsseldorf, Germany
Institute of Neuroscience and Medicine, Brain and Behaviour (INM-7), Research Center Jülich|Institute of Systems Neuroscience, Medical Faculty and University Hospital Düsseldorf, Heinrich Heine University Düsseldorf
Jülich, Germany|Düsseldorf, Germany
Matthew Cieslak, PhD
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
Theodore Satterthwaite, MD
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
The Penn Lifespan Informatics and Neuroimaging Center (PennLINC), University of Pennsylvania|Department of Psychiatry, University of Pennsylvania
Philadelphia, USA|Philadelphia, USA
Introduction:
Acquiring large-scale research neuroimaging datasets of high quality requires substantial time and resources. Hospital-acquired existing anonymized clinical neuroimaging scans represent an invaluable time and resource saving alternative that complement scientifically acquired datasets. However, due to time and resource constraints, clinical scans are often acquired with a lower quality compared to research data. To date, it remains unclear how reliable metrics extracted from these scans are. Here, we focus on diffusion MRI (dMRI) and leverage a large research dataset where two independent scans per person were acquired using a sequence similar to what may be acquired in a clinical context. To investigate how recently developed bundle segmentation methods may generalize to such clinical-grade, we evaluated the reliability of white matter (WM) bundle extraction across repeated scans and compared three commonly used reconstruction methods.
Methods:
This study used data from the Philadelphia Neurodevelopmental Cohort (Satterthwaite, 2014). After applying quality control criteria based on prior work (Roalf, 2016), we included 1,221 subjects with two dMRI scans, each using a single-shell (b=1000 s/mm2) sequence with 32 diffusion directions. All scans were pre-processed using QSIPrep (Cieslak, 2021). Subsequentially, orientation distribution functions (ODF) were reconstructed using three different methods:
1) generalized q-sampling imaging (GQI) (Yeh et al., 2010)
2) ordinary constrained spherical deconvolution (CSD) (Tournier, 2007)
3) single-shell three tissue CSD (SS3T) (Dhollander & Connelly, 2016).
The resulting ODFs were used to reconstruct 60 major WM bundles using DSI Studio's autotrack (ATK) algorithm (Yeh, 2020). Resulting bundles were warped to MNI space for comparison.
We evaluated four main outcome measures. First, we evaluated how often a given bundle could be reconstructed from the data. Second, we calculated the dice overlap between bundles from any two scans for a given method and compared within-subject to between-subject dice scores. Third, for each bundle and each reconstruction method, we calculated the discriminability (Bridgeford, 2021) of the dice scores; discriminability describes the fraction of times within-subject dice scores are larger than between-subject dice scores and provides a permutation-based statistical testing framework. Fourth and finally, we calculated probabilistic reconstruction maps for each of the bundles and compared that reconstruction map with the atlas bundle mask (Yeh, 2018) to evaluate the completeness of the reconstructed bundles.
1) generalized q-sampling imaging (GQI) (Yeh et al., 2010)
2) ordinary constrained spherical deconvolution (CSD) (Tournier, 2007)
3) single-shell three tissue CSD (SS3T) (Dhollander & Connelly, 2016).
The resulting ODFs were used to reconstruct 60 major WM bundles using DSI Studio's autotrack (ATK) algorithm (Yeh, 2020). Resulting bundles were warped to MNI space for comparison.
We evaluated four main outcome measures. First, we evaluated how often a given bundle could be reconstructed from the data. Second, we calculated the dice overlap between bundles from any two scans for a given method and compared within-subject to between-subject dice scores. Third, for each bundle and each reconstruction method, we calculated the discriminability (Bridgeford, 2021) of the dice scores; discriminability describes the fraction of times within-subject dice scores are larger than between-subject dice scores and provides a permutation-based statistical testing framework. Fourth and finally, we calculated probabilistic reconstruction maps for each of the bundles and compared that reconstruction map with the atlas bundle mask (Yeh, 2018) to evaluate the completeness of the reconstructed bundles.
Results:
Most bundles were reconstructed in all scans, with some variation by method. For GQI, 55 out of 60 bundles had reconstruction rates close to 1. This increased to 59 out of 60 bundles for both SS3T and CSD, with the left-right dentatorubrothalamic tract being the only exception.
For all bundles and all reconstruction methods, within-subject dice scores were, on average, higher than between-subject dice scores (Fig. 1A,B). Across all bundles, the within-subject dice scores were higher for CSD and SS3T compared to GQI (Fig. 1B). For the majority of bundles, discriminability values were high (> 0.9). Across all bundles, discriminability was highest for SS3T (Fig. 1C). When considering the overlap of the probabilistic maps across all scans with the atlas bundle masks (example: Fig. 2A), reconstructed bundles were most complete for CSD, closely followed by SS3T (Fig. 2B).
For all bundles and all reconstruction methods, within-subject dice scores were, on average, higher than between-subject dice scores (Fig. 1A,B). Across all bundles, the within-subject dice scores were higher for CSD and SS3T compared to GQI (Fig. 1B). For the majority of bundles, discriminability values were high (> 0.9). Across all bundles, discriminability was highest for SS3T (Fig. 1C). When considering the overlap of the probabilistic maps across all scans with the atlas bundle masks (example: Fig. 2A), reconstructed bundles were most complete for CSD, closely followed by SS3T (Fig. 2B).
Conclusions:
We found that we could reliably extract WM bundles from dMRI scans that were acquired using a simple acquisition scheme that is used in clinical practice. When comparing three different methods to reconstruct the ODFs for WM bundle reconstruction, SS3T demonstrates high discriminability and a more complete bundle reconstruction across subjects. More broadly, these results emphasize that bundle segmentation methods can be used on dMRI sequences that may be acquired in routine clinical care.
Modeling and Analysis Methods:
Diffusion MRI Modeling and Analysis 1
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
White Matter Anatomy, Fiber Pathways and Connectivity 2
Neuroinformatics and Data Sharing:
Workflows
Keywords:
Data analysis
Tractography
White Matter
WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC
Workflows
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Are you Internal Review Board (IRB) certified? Please note: Failure to have IRB, if applicable will lead to automatic rejection of abstract.
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Cieslak, M. (2021). QSIPrep: An integrative platform for preprocessing and reconstructing diffusion MRI data. Nature Methods, 18(7), 775–778.
Dhollander, T., & Connelly, A. (2016). A novel iterative approach to reap the benefits of multi-tissue CSD from just single-shell (+ b= 0) diffusion MRI data. Proc ISMRM, 24, 3010.
Roalf, D. R. (2016). The impact of quality assurance assessment on diffusion tensor imaging outcomes in a large-scale population-based cohort. NeuroImage, 125, 903–919.
Satterthwaite, T. D. (2014). Neuroimaging of the Philadelphia Neurodevelopmental Cohort. NeuroImage, 86, 544–553.
Tournier, J.-D. (2007). Robust determination of the fibre orientation distribution in diffusion MRI: Non-negativity constrained super-resolved spherical deconvolution. NeuroImage, 35(4), 1459–1472.
Yeh, F.-C. (2020). Shape analysis of the human association pathways. NeuroImage, 223, 117329.
Yeh, F.-C. (2018). Population-averaged atlas of the macroscale human structural connectome and its network topology. NeuroImage, 178, 57–68.
Yeh, F.-C. (2010). Generalized q-Sampling Imaging. IEEE Transactions on Medical Imaging, 29(9), 1626–1635.