Surface-Informed Volumetric Image Registration with MMORF
Poster No:
1489
Submission Type:
Abstract Submission
Authors:
Frederik Lange1, Rach Dawson1, Paul McCarthy1, Christoph Arthofer1, Stephen Smith1, Jesper Andersson1
Institutions:
1WIN FMRIB - University of Oxford, Oxford, UK
First Author:
Co-Author(s):
Introduction:
A shortcoming of volumetric-based registration (VBR) compared to surface-based registration (SBR) is that cortical brain regions may be aligned with less accuracy. The deformation, regularisation, and image similarity models used by most VBR tools struggle with the variability in folding patterns of the human brain. As such, compared to SBR, there remains a higher degree of variance in the location of corresponding cortical regions across individuals following VBR to a template [Robinson2014]. However, there is no straightforward way to reintegrate SBR back into volume space. In this work, we look to bridge these approaches by extending the VBR tool MMORF [Lange2024] to utilise the output of SBR to perform surface-informed volumetric registration (SI-MMORF).
Methods:
MMORF was extended to allow an arbitrary number of GIfTI (surface) image pairs, with vertex correspondence (achieved by running any SBR tool prior to MMORF), to be supplied as modalities to drive registration. The cost function implemented is the MSE of the distance between vertices in corresponding GIfTIs, with surfaces being displaced using MMORF's standard B-spline deformation model. Registration was performed between the 100 unrelated-subjects subset of the HCP dataset [VanEssen2013] and the OMM-1 template [Arthofer2024] – once using T1w and DTI images only, and once with the addition of the left and right hemisphere, FreeSurfer aligned, 32k vertex, white-matter surface GIfTIs. These GIfTIs have vertex correspondence as they were projected back into subject space following SBR. As the OMM-1 (volumetric template) does not have corresponding reference surfaces, these were created by averaging across the transformed surfaces of all HCP subjects after regular MMORF registration to the OMM-1. Note that Euclidean averaging of vertices generally underestimates cortical curvature. Performance was evaluated (Fig 1) for cortical and subcortical ROI alignment, DTI similarity, fMRI group activation, and amount of deformation.
Results:
Cortical ROI alignment by SI-MMORF noticeably improved across all metrics. Note that the difference between MMORF and SI-MMORF is greater than the difference we found between state-of-the-art (SOTA) VBR tools in a previous study [Lange 2024]. For functionally-defined MMP parcellations, the results are similar to those previously reported for the SBR tool MSMSulc [Robinson2014], and better than those previously reported for VBR tools [Coalson2018]. fMRI cluster mass (CM) was slightly lower for SI-MMORF on average, although this varied substantially depending on the contrast, and is largely attributable to a smaller number of significantly active voxels, rather than lower t-statistics. Visual inspection of the t-stat maps confirmed that the suboptimal reference surfaces (Fig 2A) caused some misalignment between the OMM-1 grey-matter mask (used in the CM calculation) and the SI-MMORF activation maps. Subcortical alignment and DTI metrics were lower for SI-MMORF, which visual inspection showed was strongly driven by the arbitrarily defined inferomedial aspects of the surfaces (Fig 2, magenta box). Aligning these vertices comes at the expense of neighbouring subcortical structures (both white and grey matter). Overall levels of volumetric distortion were similar between the two methods, but SI-MMORF showed more shape distortion (Fig 2, cyan box). This is to be expected of a method that is more closely matching regions with variable folding patterns.
Conclusions:
SI-MMORF shows promise as a method for leveraging SBR to overcome the limitations of VBR. This work highlights two main areas for improvement to address in our implementation. 1) Creating a template surface that accurately matches the volumetric anatomy. 2) Ensuring that uninformative vertices do not drive the registration. There are also several open questions that we are exploring, such as which surfaces (white, pial, mid-thickness) to use, and how to optimally weight the surface cost function.
Modeling and Analysis Methods:
Image Registration and Computational Anatomy 1
Methods Development 2
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Cortical Anatomy and Brain Mapping
Neuroinformatics and Data Sharing:
Brain Atlases
Keywords:
Spatial Normalization
Spatial Warping
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?
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.
2. Barch, D. M. (2013). Function in the human connectome: Task-fMRI and individual differences in behavior. NeuroImage, 80, 169–189.
3. Coalson, T. S. (2018). The impact of traditional neuroimaging methods on the spatial localization of cortical areas. Proceedings of the National Academy of Sciences, 115(27), E6356–E6365.
4. Destrieux, C. (2010). Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature. NeuroImage, 53(1), 1–15.
5. Irfanoglu, M. O. (2016). DR-TAMAS: Diffeomorphic Registration for Tensor Accurate Alignment of Anatomical Structures. NeuroImage, 132, 439–454.
6. Lange, F. J. (2024). MMORF—FSL’s MultiMOdal Registration Framework. Imaging Neuroscience, 2, 1–30.
7. Robinson, E. C. (2014). MSM: A new flexible framework for Multimodal Surface Matching. NeuroImage, 100, 414–426.
8. Van Essen, D. C. (2013). The WU-Minn Human Connectome Project: An overview. NeuroImage, 80, 62–79.