Micron-Scale Neural Fiber Bundle Segmentation and Reconstruction Based on PS-OCT
Poster No:
1996
Submission Type:
Abstract Submission
Authors:
Wentao Jiang1,2, Bokai Zhao1,2, Juechen Zhang1,2, Lun Sun1, Ming Song1, Tianzi Jiang1,2
Institutions:
1Institute of Automation, Chinese Academy of Sciences, Beijing, China, 2School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China
First Author:
Wentao Jiang
Institute of Automation, Chinese Academy of Sciences|School of Artificial Intelligence, University of Chinese Academy of Sciences
Beijing, China|Beijing, China
Institute of Automation, Chinese Academy of Sciences|School of Artificial Intelligence, University of Chinese Academy of Sciences
Beijing, China|Beijing, China
Co-Author(s):
Bokai Zhao
Institute of Automation, Chinese Academy of Sciences|School of Artificial Intelligence, University of Chinese Academy of Sciences
Beijing, China|Beijing, China
Institute of Automation, Chinese Academy of Sciences|School of Artificial Intelligence, University of Chinese Academy of Sciences
Beijing, China|Beijing, China
Juechen Zhang
Institute of Automation, Chinese Academy of Sciences|School of Artificial Intelligence, University of Chinese Academy of Sciences
Beijing, China|Beijing, China
Institute of Automation, Chinese Academy of Sciences|School of Artificial Intelligence, University of Chinese Academy of Sciences
Beijing, China|Beijing, China
Tianzi Jiang
Institute of Automation, Chinese Academy of Sciences|School of Artificial Intelligence, University of Chinese Academy of Sciences
Beijing, China|Beijing, China
Institute of Automation, Chinese Academy of Sciences|School of Artificial Intelligence, University of Chinese Academy of Sciences
Beijing, China|Beijing, China
Introduction:
Traditional imaging techniques for studying neural fiber bundles, such as diffusion MRI, are primarily limited to millimeter-scale resolution, making it challenging to achieve detailed characterization of neural fiber bundles (Maier-Hein et al., 2017). Polarization-sensitive Optical Coherence Tomography (PS-OCT) extends traditional OCT by incorporating polarization information, enabling high-resolution imaging and capturing tissue anisotropy, which is particularly valuable for neural fiber bundle analysis (De Boer et al., 1997; Wang et al., 2018). By integrating with block-face imaging, PS-OCT facilitates detailed 3D reconstruction of neural fiber bundles across large regions of interest (ROIs).
This study develops a multi-modal dataset based on PS-OCT and leverages a deep neural network for automated segmentation. By combining intensity and polarization features, our approach significantly improves segmentation accuracy and adaptability to complex fiber structures. The results enable precise 3D reconstruction of neural fiber bundles, providing insights into fiber topology and trajectories over large ROIs. This work demonstrates the potential of PS-OCT multi-modal imaging for advancing large-scale 3D brain reconstruction and network analysis.
This study develops a multi-modal dataset based on PS-OCT and leverages a deep neural network for automated segmentation. By combining intensity and polarization features, our approach significantly improves segmentation accuracy and adaptability to complex fiber structures. The results enable precise 3D reconstruction of neural fiber bundles, providing insights into fiber topology and trajectories over large ROIs. This work demonstrates the potential of PS-OCT multi-modal imaging for advancing large-scale 3D brain reconstruction and network analysis.
Methods:
As shown in Figure 1, this study utilized three 3D PS-OCT image blocks (256×512×256,5.9×5.9×3.5 μm³) from mouse brain specimens for segmentation. Neural fiber bundles were manually annotated by experts using ITK-SNAP (Yushkevich et al., 2006) and sliced along the y-axis into 1,536 xz-plane images for training and testing. Dual-channel multi-modal inputs were created by combining intensity and retardation images.
The nnU-Net framework was employed for automated segmentation (Isensee et al., 2021). It automatically adjusts hyperparameters based on input data and utilizes 256×256×2 multi-modal images as input. A 5-fold cross-validation strategy was employed, and performance was evaluated using the Dice coefficient (DC). Ablation experiments with single-modality inputs (intensity or retardation only) were conducted to assess the benefits of multi-modal data.
The trained nnU-Net model was further tested on a larger dataset with 20,480 images, derived from 40 3D blocks. Post-processing and stitching were performed using MATLAB, achieving 3D reconstruction of neural fiber bundles in the coronal plane, demonstrating the model's utility in capturing complex neural structures at scale.
The nnU-Net framework was employed for automated segmentation (Isensee et al., 2021). It automatically adjusts hyperparameters based on input data and utilizes 256×256×2 multi-modal images as input. A 5-fold cross-validation strategy was employed, and performance was evaluated using the Dice coefficient (DC). Ablation experiments with single-modality inputs (intensity or retardation only) were conducted to assess the benefits of multi-modal data.
The trained nnU-Net model was further tested on a larger dataset with 20,480 images, derived from 40 3D blocks. Post-processing and stitching were performed using MATLAB, achieving 3D reconstruction of neural fiber bundles in the coronal plane, demonstrating the model's utility in capturing complex neural structures at scale.
·Figure 1: Flowchart of Neural Fiber Bundle Segmentation and 3D Reconstruction Based on PS-OCT
Results:
This study evaluated neural fiber bundle segmentation using Dice coefficient (DC). In 5-fold cross-validation, the nnU-Net model with dual-channel multi-modal inputs achieved an average DC of 0.8596. Ablation experiments showed that single-modal inputs using intensity and retardation images resulted in DC scores of 0.8411 and 0.8092, respectively, emphasizing the accuracy and robustness of multi-modal inputs.
A traditional U-Net (Ronneberger et al., 2015) achieved a lower DC of 0.8369, confirming the advantage of nnU-Net with adaptive hyperparameter tuning. On a larger dataset, nnU-Net effectively segmented neural fiber bundles of varying sizes, including the corpus callosum (CC) and fibers near the hippocampal formation (HPF), as shown in Figure 2, demonstrating strong generalization and robustness for neural fiber analysis and brain network construction.
A traditional U-Net (Ronneberger et al., 2015) achieved a lower DC of 0.8369, confirming the advantage of nnU-Net with adaptive hyperparameter tuning. On a larger dataset, nnU-Net effectively segmented neural fiber bundles of varying sizes, including the corpus callosum (CC) and fibers near the hippocampal formation (HPF), as shown in Figure 2, demonstrating strong generalization and robustness for neural fiber analysis and brain network construction.
·Figure 2: Example of Neural Fiber Bundle Segmentation Results (Corpus Callosum and Fibers Near the Hippocampal Formation)
Conclusions:
This study leverages PS-OCT imaging in combination with nnU-Net to achieve automated neural fiber bundle segmentation and 3D reconstruction. Multi-modal inputs (intensity and retardation images) significantly improved accuracy, achieving a DC of 0.8596, outperforming both single-modal inputs and U-Net.
The model demonstrated strong generalization on larger datasets, accurately segmenting diverse fiber structures, including the corpus callosum and fibers near the hippocampal formation. This demonstrates PS-OCT's potential for large-scale neural fiber analysis and brain network research.
The model demonstrated strong generalization on larger datasets, accurately segmenting diverse fiber structures, including the corpus callosum and fibers near the hippocampal formation. This demonstrates PS-OCT's potential for large-scale neural fiber analysis and brain network research.
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
White Matter Anatomy, Fiber Pathways and Connectivity 2
Novel Imaging Acquisition Methods:
Optical coherence tomography (OCT) 1
Keywords:
Data analysis
White Matter
Other - Polarization-sensitive Optical Coherence Tomography
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:
Provide references using APA citation style.
Isensee, F., Jaeger, P. F., Kohl, S. A. A., Petersen, J., & Maier-Hein, K. H. (2021). nnU-Net: A self-configuring method for deep learning-based biomedical image segmentation. Nature Methods, 18(2), 203–211.
Maier-Hein, K. H., Neher, P. F., Houde, J.-C., Côté, M.-A., Garyfallidis, E., Zhong, J., Chamberland, M., Yeh, F.-C., Lin, Y.-C., Ji, Q., Reddick, W. E., Glass, J. O., Chen, D. Q., Feng, Y., Gao, C., Wu, Y., Ma, J., He, R., Li, Q., … Descoteaux, M. (2017). The challenge of mapping the human connectome based on diffusion tractography. Nature Communications, 8(1), 1349.
Ronneberger, O., Fischer, P., & Brox, T. (2015). U-Net: Convolutional Networks for Biomedical Image Segmentation. In N. Navab, J. Hornegger, W. M. Wells, & A. F. Frangi (Eds.), Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015 (Vol. 9351, pp. 234–241). Springer International Publishing.
Wang, H., Magnain, C., Wang, R., Dubb, J., Varjabedian, A., Tirrell, L. S., Stevens, A., Augustinack, J. C., Konukoglu, E., Aganj, I., Frosch, M. P., Schmahmann, J. D., Fischl, B., & Boas, D. A. (2018). as-PSOCT: Volumetric microscopic imaging of human brain architecture and connectivity. NeuroImage, 165, 56–68.
Yushkevich, P. A., Piven, J., Hazlett, H. C., Smith, R. G., Ho, S., Gee, J. C., & Gerig, G. (2006). User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability. NeuroImage, 31(3), 1116–1128.