Attention-gated Convolutional Neural Network for Automated Segmentation of Fetal Subplate from MRI.
Poster No:
1656
Submission Type:
Late-Breaking Abstract Submission
Authors:
Andrea Gondova1, Milton O. Candela-Leal1, Hyuk Jin Yun1, Sungmin You1, Seungyoon Jeong1, Marisol Aguilar1, P. Ellen Grant1, Kiho Im1
Institutions:
1Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital, Boston, MA
First Author:
Andrea Gondova
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Co-Author(s):
Milton O. Candela-Leal
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Hyuk Jin Yun
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Sungmin You
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Seungyoon Jeong
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Marisol Aguilar
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
P. Ellen Grant
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Kiho Im
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital
Boston, MA
Late Breaking Reviewer(s):
Introduction:
The subplate (SP) is a transient fetal brain structure implicated in numerous developmental processes, including neuronal migration, circuit formation, and early cortical activity, with its alterations linked to neurodevelopmental disorders (Kostović, 2020). While recent advances in fetal MRI enable in vivo SP visualization, automated segmentation ‒ crucial for large-scale, non-invasive studies ‒ remains underexplored compared to other brain structures like the cortical plate (CP) (Dou et al., 2020; Uus et al., 2023). To address this gap, we extend a validated attention-gated U-Net model (Hong et al., 2020), originally developed for CP segmentation, to incorporate SP to facilitate a more comprehensive fetal brain analysis.
Methods:
We analyzed a retrospective multi-site cohort of T2-weighted fetal brain MRI from 134 typically developing fetuses (GW: 27.32 ± 2.67, range: 21.86–31.86). Data processing included brain masking, slice-to-volume registration, alignment to a 31-week template (You et al., 2024), and N4 bias field correction to mitigate scanner variations (Tutison et al., 2010). Fourteen cases were held out for testing, while 120 were used for training and validation (Figure 1a,b). Our approach consisted of an ensemble of three 2D Attention Gated U-Net models trained on axial, coronal, and sagittal planes, with multi-view aggregation and test-time augmentation (MVA-TTA) to improve segmentation stability. A hybrid loss function combining Dice and focal loss was used to enhance boundary delineation (Hong et al., 2020)(Figure 1c).
Results:
Our model demonstrated high segmentation performance, achieving a global Dice score of 0.98 ± 0.015 (mean ± std), Hausdorff distance of 5.36 ± 2.406 voxels, and 1.29 ± 0.929% volumetric change in the test set. Performance was consistent across tissue types. We observed a slight (non-significant) decline increasing gestational age, likely due to evolving tissue contrast and morphology complexity (Figure 2a). Qualitative evaluations showed most segmentation errors at the CP/SP boundary (Figure 2b). Despite these, our method provided sufficient accuracy for volumetric and morphometric SP analyses, significantly reducing the need for extensive manual correction in follow-up studies (data not shown).
Conclusions:
We present an automated deep-learning approach for automated SP segmentation in fetal MRI, expanding existing method for large-scale studies of typical and atypical SP development. While minor manual correction remains necessary, particularly in low-quality scans or complex morphologies at later gestational stages, future improvements could incorporate automated refinement techniques such as conditional random fields (Chen et al., 2021) or hybrid segmentation approaches (Gaset et al., 2024). By reducing reliance on labor-intensive manual segmentation, this work will allow future work that could enhance the study of in utero brain development, advance our understanding of early brain development and detect early biomarker for various clinical populations with adverse neurodevelopmental outcomes.
Lifespan Development:
Normal Brain Development: Fetus to Adolescence 2
Modeling and Analysis Methods:
Methods Development
Segmentation and Parcellation 1
Keywords:
Cortical Layers
Development
Machine Learning
PEDIATRIC
Segmentation
STRUCTURAL MRI
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.
Diogo, M. C., Prayer, D., Gruber, G. M., Brugger, P. C., Stuhr, F., Weber, M., Bettelheim, D., & Kasprian, G. (2019). Echo-planar FLAIR Sequence Improves Subplate Visualization in Fetal MRI of the Brain. Radiology, 292(1), 159–169
Dou, H., Karimi, D., Rollins, C.K., Ortinau, C.M., Vasung, L., Velasco-Annis, C., Ouaalam, A., Yang, X., Ni, D., & Gholipour, A. (2020). A Deep Attentive Convolutional Neural Network for Automatic Cortical Plate Segmentation in Fetal MRI. IEEE Transactions on Medical Imaging, 40, 1123-1133.
Gaser Ch., Dahnke R., (2024) CAT faces Python – Combining Deep-Learning with Traditional Segmentation. OHBM 2024 Annual Meeting, Seoul, South Korea.
Hong, J., Yun, H.J., Park, G., Kim, S., Laurentys, C.T., Siqueira, L.C., Tarui, T., Rollins, C.K., Ortinau, C.M., Grant, P.E., Lee, J., & Im, K. (2020). Fetal Cortical Plate Segmentation Using Fully Convolutional Networks With Multiple Plane Aggregation. Frontiers in Neuroscience, 14.
Kostović, I. (2020). The enigmatic fetal subplate compartment forms an early tangential cortical nexus and provides the framework for construction of cortical connectivity. Progress in Neurobiology, 194.
Tustison, N., Avants, B.B., Cook, P.A., Zheng, Y., Egan, A., Yushkevich, P., & Gee, J.C. (2010). N4ITK: Improved N3 Bias Correction. IEEE Transactions on Medical Imaging, 29, 1310-1320.
Uus, A. U., Kyriakopoulou, V., Makropoulos, A., Fukami-Gartner, A., Cromb, D., Davidson, A., Cordero-Grande, L., Price, A. N., Grigorescu, I., Williams, L. Z. J., Robinson, E. C., Lloyd, D., Pushparajah, K., Story, L., Hutter, J., Counsell, S. J., Edwards, A. D., Rutherford, M. A., Hajnal, J. V., & Deprez, M. (2023). BOUNTI: Brain vOlumetry and aUtomated parcellatioN for 3D feTal MRI. eLife 12:RP88818.
You, S., Barba, A.D., Tamayo, V.C., Yun, H.J., Yang, E., Grant, P.E., Im, K., Wu, Z., Zhao, F., Leon, D., Tamayo, C., Yun, H., & Grant, P.E. (2024). Automatic cortical surface parcellation in the fetal brain using attention-gated spherical U-net. Frontiers in Neuroscience, 18.