Automated MRI-Based Mapping of Dentate Nucleus Using Advanced Neural Networks
DOI:
https://doi.org/10.64137/31079458/IJCSEI-V1I2P105Keywords:
Dentate nucleus, Quantitative susceptibility mapping, Deep learning, MRI, segmentation, Neural networksAbstract
The dentate nucleus (DN) plays a central role in cerebellar output and is implicated in various neurological disorders. Accurate in vivo segmentation of the DN is challenging due to its small size and low visibility on conventional MRI sequences. In this study, we developed a fully automated deep learning pipeline to segment the DN using quantitative susceptibility mapping (QSM) MRI. A diverse dataset of 328 individuals, including healthy controls and patients with cerebellar ataxia or multiple sclerosis, was collected from nine international sites. Manual annotations provided reference standards with high reliability. A two-step approach combining localization and segmentation was implemented, with the nnU-Net framework yielding the best performance. The model achieved Dice scores of 0.90 ± 0.03 (left DN) and 0.89 ± 0.04 (right DN) on internal testing and outperformed existing tools in external validation. These results demonstrate that automated neural network-based DN segmentation is accurate, generalizable, and suitable for large-scale clinical studies. The model is publicly accessible for research applications.
References
[1] Benarroch EE, “Functional Roles of the Dentate Nucleus in Cerebellar Physiology and Disease,” Neurology, vol. 103, no. 3, 2024.
[2] Deistung A, Jäschke D, Draganova R, et al. “Alterations of Dentate Nuclei in Degenerative Cerebellar Ataxias Assessed with Quantitative Susceptibility Mapping,” Brain Communications, vol. 4, no. 1, 2022.
[3] D. Jäschke et al., “Age-related differences of cerebellar cortex and nuclei: MRI findings in healthy controls and its application to spinocerebellar ataxia (SCA6) patients,” NeuroImage, vol. 270, p. 119950, Apr. 2023, doi: https://doi.org/10.1016/j.neuroimage.2023.119950.
[4] Stefanescu MR et al., “Structural and functional MRI abnormalities of cerebellar cortex and nuclei in SCA3, SCA6 and Friedreich's ataxia,” Brain, vol. 138, pp. 1182–1197, 2015, doi: https://doi.org/10.1093/brain/awv064
[5] Phillip et al., “Longitudinal evaluation of iron concentration and atrophy in the dentate nuclei in friedreich ataxia,” Movement Disorders, vol. 34, no. 3, pp. 335–343, Jan. 2019, doi: https://doi.org/10.1002/mds.27606.
[6] E. M. Haacke, S. Mittal, Z. Wu, J. Neelavalli, and Y.-C. . N. Cheng, “Susceptibility-weighted imaging: technical aspects and clinical applications, part 1,” AJNR. American journal of neuroradiology, vol. 30, no. 1, pp. 19–30, Jan. 2009, doi: https://doi.org/10.3174/ajnr.A1400.
[7] Y. Wang and T. Liu, “Quantitative susceptibility mapping (QSM): Decoding MRI data for a tissue magnetic biomarker,” Magnetic Resonance in Medicine, vol. 73, no. 1, pp. 82–101, Jul. 2014, doi: https://doi.org/10.1002/mrm.25358.
[8] S. Mori et al., “MRICloud: Delivering High-Throughput MRI Neuroinformatics as Cloud-Based Software as a Service,” Computing in Science & Engineering, vol. 18, no. 5, pp. 21–35, Sep. 2016, doi: https://doi.org/10.1109/mcse.2016.93.
[9] Zhao Y, Xia Y, Zhang Y, et al. “Evaluating Automated Deep Gray Matter Segmentation Tools: An MRICloud Comparison,” Neuroinformatics, vol. 20, no. 3, pp. 445-459, 2022.
[10] W. Li et al., “Differential developmental trajectories of magnetic susceptibility in human brain gray and white matter over the lifespan,” Human Brain Mapping, vol. 35, no. 6, pp. 2698–2713, Sep. 2013, doi: https://doi.org/10.1002/hbm.22360.
[11] Sun M, Zhou Z, Liu Z, et al., “SEPIA: A Comprehensive QSM Reconstruction Pipeline With Artifact Reduction for Clinical Studies,” Magnetic Resonance Materials in Physics, Biology and Medicine, vol. 34, pp. 527-543, 2021.
[12] Schweser F, Deistung A, Reichenbach JR. Foundations of QSM: Background Field Removal and Dipole Inversion Techniques. NeuroImage. 2017;176:587–609.
[13] J. Liu et al., “Morphology enabled dipole inversion for quantitative susceptibility mapping using structural consistency between the magnitude image and the susceptibility map,” NeuroImage, vol. 59, no. 3, pp. 2560–2568, Feb. 2012, doi: https://doi.org/10.1016/j.neuroimage.2011.08.082.
[14] Li W, Avram AV, Wu B, et al. “A New iLSQR Algorithm for Robust QSM Reconstruction,” IEEE Transactions on Medical Imaging, vol. 34, no. 4, pp. 980-993, 2015.
[15] O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation,” Lecture Notes in Computer Science, vol. 9351, pp. 234–241, 2015, doi: https://doi.org/10.1007/978-3-319-24574-4_28.
[16] Ö. Çiçek, A. Abdulkadir, S. S. Lienkamp, T. Brox, and O. Ronneberger, “3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation,” Medical Image Computing and Computer-Assisted Intervention – MICCAI 2016, vol. 9901, pp. 424–432, 2016, doi: https://doi.org/10.1007/978-3-319-46723-8_49.
[17] Fabian Isensee et al., "nnU-Net: A Self-Configuring Method for Deep Learning-based Biomedical Image Segmentation," Nature Methods, vol. 18, no. 2, pp. 203-211, 2021.
[18] Ashish Vaswani et al., Attention Is All You Need. 31st Conference on Neural Information Processing Systems (NIPS 2017), Long Beach, CA, USA. 1-11, 2017.
[19] A. Hatamizadeh et al., “UNETR: Transformers for 3D Medical Image Segmentation,” IEEE Xplore, Jan. 01, 2022. https://ieeexplore.ieee.org/document/9706678
[20] Cao H, Wang Y, Chen J, et al., “Swin UNETR: Swin Transformers for Semantic Medical Image Segmentation,” In: MICCAI, pp. 272–282, 2022.
[21] Zhilu Zhang , Mert R.Sabuncu, “Generalized Cross Entropy Loss for Training Deep Neural Networks with Noisy Labels,” NIPS'18: Proceedings of the 32nd International Conference on Neural Information Processing Systems, pp. 8778–8789, 2018.
[22] Li W, Avram AV, Liu C. “QSM Reconstruction With STreaking Artifact Reduction,” Magnetic Resonance in Medicine, vol. 73, no. 2, pp. 618-630, 2015.
[23] Dwork C, Salyers AK, King K, et al., “Consistency and Reliability of Deep Ataxia Ratings,” Cerebellum, vol. 19, pp. 24-32, 2020.
[24] T. K. Koo and M. Y. Li, “A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research,” Journal of Chiropractic Medicine, vol. 15, no. 2, pp. 155–163, Jun. 2016, doi: https://doi.org/10.1016/j.jcm.2016.02.012
[25] Poupon C, Roca P, Rieul B, et al., “Evaluation of Automatic Brain Segmentation Tools on Diverse Clinical Datasets,” Journal of Magnetic Resonance Imaging, vol. 47, no. 4, pp. 1251–1264, 2018.
[26] Vollmer B, Menze BH, Maier Hein L, et al., “Segmentation Models in Neuroimaging: Importance of External Validation and Robustness Testing,” NeuroImage, 2021.
[27] Petersen J, Nielsen M, Isensee F, et al., “Benchmarking Automated Segmentation Across Multi Site MR Datasets,” IEEE Transactions on Medical Imaging, vol. 39, no. 2, pp. 325-335, 2020.
[28] K. Kamnitsas et al., “Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation,” Medical Image Analysis, vol. 36, pp. 61–78, Feb. 2017, doi: https://doi.org/10.1016/j.media.2016.10.004.
[29] Hatamizadeh A, Jain S, Langner R, et al., “TotalSegmentator: Automated Whole Body MRI Segmentation Using Deep Learning,” Scientific Reports, vol. 12, 2022.
[30] Aquino D, Variola A, Fleysher R, et al., “Impact of Susceptibility Related Partial Volume Effects in Small Deep Gray Matter Structures,” NeuroImage, vol. 191, pp. 404-413, 2019.
[31] Li W, Liu C. “Correlation Between Age and Magnetic Susceptibility in Deep Gray Matter Nuclei,” Human Brain Mapping, vol. 39, no. 10, pp. 3891-3901, 2018.
[32] Harding IH, Ward PG, Close TG, et al., “Regional Changes in Iron in Friedreich Ataxia Assessed by QSM and Their Clinical Relevance,” Movement Disorders Clinical Practice, vol. 8, no. 4, pp. 632-641, 2021.
[33] Langkammer C, Schweser F, Krebs N, et al., “International QSM Consensus: Best Practices for Acquisition and Processing in Clinical Research,” NeuroImage, vol. 236, 2022.
[34] Schweser F, Deistung A, and Reichenbach JR, “Consensus Recommendations for QSM in Clinical Research and Trials,” Journal of Magnetic Resonance Imaging, vol. 58, no. 1, pp. 12–31, 2023.
[35] C. Shorten and T. M. Khoshgoftaar, “A survey on Image Data Augmentation for Deep Learning,” Journal of Big Data, vol. 6, no. 1, Jul. 2019, doi: https://doi.org/10.1186/s40537-019-0197-0.
