Shape Matters: Correlates of Subcortical Anatomy with Huntington’s Disease Progression
Poster No:
250
Submission Type:
Late-Breaking Abstract Submission
Authors:
Mohsen Ghofrani-Jahromi1, Susmita Saha1, Adeel Razi2, Pubu Abeyasinghe2, Govinda Poudel3, Jane Paulsen4, Sarah Tabrizi5, Nellie Georgiou-Karistianis6
Institutions:
1Monash University, Melbourne, Victoria, 2Monash University, Melbourne, VIC, 3Mary Mackillop Institute, Melbourne, VIC, 4University of Wisconsin-Madison, Madison, WI, 5University College London, London, United Kingdom, 6Monash University, Clayton, Victoria
First Author:
Co-Author(s):
Introduction:
Huntington's disease (HD) is a progressive neurodegenerative disorder, where clinical trials rely on sensitive biomarkers to stratify participants and to monitor therapeutic effects. Despite evidence of group differences in striatal morphometry among premanifest (pre-HD), symptomatic HD (symp-HD) and healthy controls, current models of HD progression do not incorporate shape information. Novel deep learning methods can derive effective descriptors of anatomical shape from subcortical brain structures strongly associated with disease progression across HD continuum. Integrating these anatomical features into progression models could enhance biomarker sensitivity and improve clinical assessments.
Methods:
We incorporated data from four cohorts: TRACK-HD, TrackON-HD, PREDICT-HD, and IMAGE-HD. Each dataset featured repeated visits, during which structural MRI was conducted (2,932 brain scans in 615 HD individuals).
We used FreeSurfer for MRI pre-processing and segmentation. We investigated whether PointNet, a neural network for processing point clouds, is capable of deriving shape descriptors from each of the segmented structures that could estimate the severity of HD progression as measured by PINHD, a prognostic score for HD that encompasses genetic, cognitive, and motor assessments.
A PointNet processes a set of N unordered points on the surface of a segmented structure, treating each point in the 3D space identically and independently in two steps:
Mapping to a high-dimensional space via a fully connected network that take P(N×3) as input and give H(N×F) as output, where F is the length of the shape descriptor vector.
A symmetric operator, e.g. a max pooling layer, which is essentially invariant to the permutation of its inputs is applied column-wise on the matrix H(N×F) to yield F shape descriptors.
Principal component analysis (PCA) was used for dimensionality reduction and to compare quantitatively the effectiveness of stacked shape descriptors and conventionally used volumetric measures (mm3) in reflecting disease progression in the unseen test set.
We used FreeSurfer for MRI pre-processing and segmentation. We investigated whether PointNet, a neural network for processing point clouds, is capable of deriving shape descriptors from each of the segmented structures that could estimate the severity of HD progression as measured by PINHD, a prognostic score for HD that encompasses genetic, cognitive, and motor assessments.
A PointNet processes a set of N unordered points on the surface of a segmented structure, treating each point in the 3D space identically and independently in two steps:
Mapping to a high-dimensional space via a fully connected network that take P(N×3) as input and give H(N×F) as output, where F is the length of the shape descriptor vector.
A symmetric operator, e.g. a max pooling layer, which is essentially invariant to the permutation of its inputs is applied column-wise on the matrix H(N×F) to yield F shape descriptors.
Principal component analysis (PCA) was used for dimensionality reduction and to compare quantitatively the effectiveness of stacked shape descriptors and conventionally used volumetric measures (mm3) in reflecting disease progression in the unseen test set.
Results:
From the eight investigated subcortical structures, shape descriptors from the lateral ventricle, thalamus, caudate, putamen, pallidum, and accumbens showed strong association with HD progression, with hippocampus and amygdala having minor association. Figure 1 illustrates the estimated versus the actual values of PIN scores for each PointNet per structure.
The Spearman's rank correlation between one dimensional principal component and HD Integrated Staging (HD-ISS) was higher for shape descriptors compared to volumetric measures (ρ = 0.72 vs ρ = 0.45) indicating better capture of disease progression in this feature space.
The Spearman's rank correlation between one dimensional principal component and HD Integrated Staging (HD-ISS) was higher for shape descriptors compared to volumetric measures (ρ = 0.72 vs ρ = 0.45) indicating better capture of disease progression in this feature space.
·Two-dimensional visualization of feature spaces using PCA for: (A) volumetric measures versus (B) shape descriptors from the putamen, pallidum, caudate, lateral ventricle, thalamus, and accumbens.
·A comparison between the capability of shape descriptors derived from different subcortical structures in estimating disease progression measured by the concurrent PIN score.
Conclusions:
We demonstrated the feasibility of training a discriminative DL architecture that derived meaningful descriptors from brain subcortical structures to associate shape with HD progression. Our findings confirmed that descriptors derived from the putamen, caudate, pallidum, and the lateral ventricle exhibited strong predictive associations with HD progression.
Moreover, shape descriptors could represent the trajectory of disease progression from HD-ISS stage 0 to 3 more granularly, when compared to commonly used volumetric measures.
Moreover, shape descriptors could represent the trajectory of disease progression from HD-ISS stage 0 to 3 more granularly, when compared to commonly used volumetric measures.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 1
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Keywords:
Other - Huntington’s Disease
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.
Tabrizi, S. J., Schobel, S., Gantman, E. C., Mansbach, A., Borowsky, B., Konstantinova, P., ... & Sampaio, C. (2022). A biological classification of Huntington's disease: the Integrated Staging System. The Lancet Neurology, 21(7), 632-644.