TY - GEN
T1 - Construction of Spatiotemporal Infant Cortical Surface Functional Templates
AU - the UNC/UMN Baby Connectome Project Consortium
AU - Huang, Ying
AU - Wang, Fan
AU - Wu, Zhengwang
AU - Chen, Zengsi
AU - Zhang, Han
AU - Wang, Li
AU - Lin, Weili
AU - Shen, Dinggang
AU - Li, Gang
N1 - Funding Information:
Acknowledgments. This work was partially supported by NIH grants (MH116225 and MH117943). This work also utilizes approaches developed by an NIH grant (1U01MH110274) and the efforts of the UNC/UMN Baby Connectome Project Consortium.
PY - 2020
Y1 - 2020
N2 - Infant cortical surface templates play an essential role in spatial normalization of cortical surfaces across individuals in pediatric neuroimaging analysis. However, existing infant surface templates have two major limitations in functional MRI analysis. First, they are constructed by co-registration of cortical surfaces based on structural attributes, which cannot lead to accurate functional alignment, due to the highly variable relationship between cortical folds and functions. Second, they are constructed by simply averaging co-registered cortical attributes, which is sensitive to registration errors and lead to blurred attribute patterns on templates, thus deteriorating the accuracy in spatial normalization. Therefore, construction of infant cortical functional templates encoding sharp functional architectures is critical for infant fMRI analysis. To this end, we construct the first set of spatiotemporal infant cortical surface functional templates using Wasserstein barycenter and a state-of-the-art functional feature, namely the gradient density of functional connectivity. To address the first issue, we leverage functional gradient density to drive surface registration to improve inter-individual functional correspondences. To address the second issue, we compute templates based on the Wasserstein barycenter of functional gradient density maps across individuals. The motivation is that Wasserstein barycenter represents a meaningful mean under the Wasserstein distance metric, which takes into account the alignment of local spatial distribution of cortical attributes and thus is robust to registration errors, leading to sharp and detailed patterns on templates. Experiments on a dataset with 207 fMRI scans between 0 and 2 years of age show the validity and accuracy of our constructed infant cortical functional templates.
AB - Infant cortical surface templates play an essential role in spatial normalization of cortical surfaces across individuals in pediatric neuroimaging analysis. However, existing infant surface templates have two major limitations in functional MRI analysis. First, they are constructed by co-registration of cortical surfaces based on structural attributes, which cannot lead to accurate functional alignment, due to the highly variable relationship between cortical folds and functions. Second, they are constructed by simply averaging co-registered cortical attributes, which is sensitive to registration errors and lead to blurred attribute patterns on templates, thus deteriorating the accuracy in spatial normalization. Therefore, construction of infant cortical functional templates encoding sharp functional architectures is critical for infant fMRI analysis. To this end, we construct the first set of spatiotemporal infant cortical surface functional templates using Wasserstein barycenter and a state-of-the-art functional feature, namely the gradient density of functional connectivity. To address the first issue, we leverage functional gradient density to drive surface registration to improve inter-individual functional correspondences. To address the second issue, we compute templates based on the Wasserstein barycenter of functional gradient density maps across individuals. The motivation is that Wasserstein barycenter represents a meaningful mean under the Wasserstein distance metric, which takes into account the alignment of local spatial distribution of cortical attributes and thus is robust to registration errors, leading to sharp and detailed patterns on templates. Experiments on a dataset with 207 fMRI scans between 0 and 2 years of age show the validity and accuracy of our constructed infant cortical functional templates.
KW - Functional templates
KW - Infant
KW - Wasserstein barycenter
UR - http://www.scopus.com/inward/record.url?scp=85092720883&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85092720883&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-59728-3_24
DO - 10.1007/978-3-030-59728-3_24
M3 - Conference contribution
AN - SCOPUS:85092720883
SN - 9783030597276
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 238
EP - 248
BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2020 - 23rd International Conference, Proceedings
A2 - Martel, Anne L.
A2 - Abolmaesumi, Purang
A2 - Stoyanov, Danail
A2 - Mateus, Diana
A2 - Zuluaga, Maria A.
A2 - Zhou, S. Kevin
A2 - Racoceanu, Daniel
A2 - Joskowicz, Leo
PB - Springer Science and Business Media Deutschland GmbH
T2 - 23rd International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2020
Y2 - 4 October 2020 through 8 October 2020
ER -