Towards a Comprehensive Toolkit for Human Visual Cortex Parcellation
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:
1531
Submission Type:
Abstract Submission
Authors:
Fernanda Ribeiro1, Torin Bambridge-Lozan2, Noah Benson3, D. Samuel Schwarzkopf4, Martin Hebart1, Alexander Puckett5, Steffen Bollmann2
Institutions:
1Justus-Liebig University Giessen, Giessen, Hessen, 2The University of Queensland, Brisbane, Queensland, 3University of Washington, Seattle, WA, 4University of Auckland, Auckland, Auckland, 5University of Technology Sydney, Sydney, New South Wales
First Author:
Co-Author(s):
Introduction:
The visual system comprises several functionally specialized cortical areas, nearly all of which contain representations of the visual field. In these areas, adjacent neurons represent adjacent retinal locations. These retinotopic maps are typically defined in polar coordinates, resulting in two orthogonal coordinate maps: polar angle and eccentricity. While the retinotopic organization of early visual areas (V1, V2, and V3) in the human cortex is generally assumed to be organized according to a universal topological template that is similar across people (Figure 1a), recent investigations have revealed compelling evidence of interindividual topological differences [1]. These differences cast doubt on the traditional template of early visual cortex (Figure 1b). Therefore, we propose a unified, automated solution for retinotopic mapping and visual cortex parcellation that is based only on anatomical data that can be derived from a T1-weighted image and that is not dependent on any single template of retinotopic organization. We show that our integrated solution predicts accurate retinotopic maps, and corresponding visual field sign maps reveal the boundaries between early visual areas.
·Figure 1. Interindividual variability in retinotopic mapping. a. The canonical model for the polar angle representation in V1, V2, and V3. b. Three maps that deviate from the canonical model.
Methods:
Our toolbox integrates (1) standard neuroimaging software (FreeSurfer 7.3.2 and Connectome Workbench 1.5.0) for anatomical MRI data preprocessing, (2) a deep-learning model for predicting retinotopic maps at the individual level [2], and (3) an efficient implementation of the visual field sign analysis for early visual areas parcellation (Figure 2a). These components are packaged into Docker and Singularity software containers, which can be easily downloaded and are available on NeuroDesk [3]. Details for these individual components follow:
(1) Input data preprocessing: Structural MRI data from the Human Connectome Project Retinotopy dataset [4] is preprocessed with FreeSurfer recon-all for 'white' and 'pial' surface tessellation. Then, we generate the midthickness surface, for which we determine the curvature map. These curvature maps are then resampled to the standard HCP '32k_fs_LR' surface space using Connectome Workbench.
(2) Retinotopic maps generation: We leverage deepRetinotopy [2], a geometric deep learning model for retinotopic mapping, to predict individual-specific retinotopic maps from curvature maps. To ensure generalizability, we retrained deepRetinotopy to predict retinotopic maps solely from curvature maps generated with FreeSurfer.
(3) Visual field sign analysis: This analysis combines both the polar angle and eccentricity maps to detect whether deforming the visual field onto the cortical surface requires a reflection of the visual field or not. Because adjacent visual areas have distinct field signs, visual area boundaries can be clearly visualized using the field sign [5]. In our implementation, we project the surface mesh representation in 'brain space' to visual field coordinate space, where we compute the normal vector of each surface mesh triangle. The sign of the normal vector reveals whether each visual area contains a reflected (negative) or unreflected (positive) deformation of the visual field.
(1) Input data preprocessing: Structural MRI data from the Human Connectome Project Retinotopy dataset [4] is preprocessed with FreeSurfer recon-all for 'white' and 'pial' surface tessellation. Then, we generate the midthickness surface, for which we determine the curvature map. These curvature maps are then resampled to the standard HCP '32k_fs_LR' surface space using Connectome Workbench.
(2) Retinotopic maps generation: We leverage deepRetinotopy [2], a geometric deep learning model for retinotopic mapping, to predict individual-specific retinotopic maps from curvature maps. To ensure generalizability, we retrained deepRetinotopy to predict retinotopic maps solely from curvature maps generated with FreeSurfer.
(3) Visual field sign analysis: This analysis combines both the polar angle and eccentricity maps to detect whether deforming the visual field onto the cortical surface requires a reflection of the visual field or not. Because adjacent visual areas have distinct field signs, visual area boundaries can be clearly visualized using the field sign [5]. In our implementation, we project the surface mesh representation in 'brain space' to visual field coordinate space, where we compute the normal vector of each surface mesh triangle. The sign of the normal vector reveals whether each visual area contains a reflected (negative) or unreflected (positive) deformation of the visual field.
Results:
Our toolkit can generate detailed, individual-specific retinotopic maps (Figure 2b). Moreover, with polar angle and eccentricity (not shown) maps, our toolkit generates visual field sign representations where boundaries between early visual areas are unambiguous.
·Figure 2. A comprehensive toolkit for human visual cortex parcellation. a. Toolkit components. b. Empirical and predicted polar angle maps with corresponding visual field sign maps.
Conclusions:
These results demonstrate the potential of our open-source toolbox (https://github.com/felenitaribeiro/deepRetinotopy_TheToolbox) for individual-specific visual cortex parcellation. This toolkit is the first step towards a fully automated approach for early visual area parcellation; for example, the predicted retinotopic maps can be supplemented with empirical data to improve parcellation results using other available tools [6]. Future work will seek an effective solution for integrating existing tools for visual cortex parcellation.
Modeling and Analysis Methods:
Methods Development 1
Segmentation and Parcellation
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Anatomy and Functional Systems
Cortical Anatomy and Brain Mapping 2
Keywords:
Computing
Cortex
FUNCTIONAL MRI
Machine Learning
Open-Source Software
STRUCTURAL MRI
Vision
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] Ribeiro, F. L., Bollmann, S. & Puckett, A. M. Predicting the retinotopic organization of human visual cortex from anatomy using geometric deep learning. NeuroImage 118624 (2021).
[3] Renton, A.I., Dao, T.T., et al., Neurodesk: An accessible, flexible, and portable data analysis environment for reproducible neuroimaging, Nature Methods (2024).
[4] Benson, N.C. et al. The Human Connectome Project 7 Tesla retinotopy dataset: Description and population receptive field analysis. Journal of Vision 18:23 (2018).
[5] Sereno, M.I., McDonald, C.T., Allman, J.M. Analysis of retinotopic maps in extrastriate cortex. Cerebral Cortex 4:601-620 (1994).
[6] Benson, N.C., Winawer, J. Bayesian analysis of retinotopic maps. eLife (2018)