Deeply Supervised 3D U-Net with Auxiliary Classification for Cerebral Aneurysms Segmentation

Authors

  • Mario Mata-Castillo Universidad Autónoma de Ciudad Juárez, Ciudad Juárez, Mexico
  • Nelly Gordillo-Castillo Universidad Autónoma de Ciudad Juárez, Ciudad Juárez, Mexico
  • Boris Mederos Universidad Autónoma de Ciudad Juárez, Ciudad Juárez, Mexico
  • Dora-Luz Flores Universidad Autónoma de Baja California, Mexicali, Mexico https://orcid.org/0000-0002-2623-3772
  • Vicente García Universidad Autónoma de Ciudad Juárez, Ciudad Juárez, Mexico https://orcid.org/0000-0003-2820-2918
  • José Díaz-Román Universidad Autónoma de Ciudad Juárez, Ciudad Juárez, Mexico https://orcid.org/0000-0002-8246-6562

DOI:

https://doi.org/10.3991/ijoe.v22i10.62047

Keywords:

segmentation, deep supervision, magnetic resonance angiography, cerebral aneurysm, ; Deep Learning

Abstract


Cerebral aneurysms are dilations of intracranial arteries whose early detection is essential for treatment planning and rupture prevention. Automated aneurysm segmentation remains challenging because of severe class imbalance and the small size of many lesions. This study presents a 3D U-Net incorporating deep supervision and an auxiliary classification branch, trained on the public ADAM dataset using skeleton-guided patch extraction. Fixed-size patches (64 × 64 × 64) were extracted from volumes resampled to 0.4 mm isotropic resolution, and performance was evaluated using stratified five-fold cross-validation. The proposed model achieved a Dice coefficient of 0.425 (95% CI: 0.362–0.488) and a false-positive (FP) patch rate of 0.026 (95% CI: 0.017–0.035) on the test set. For aneurysms larger than 2.5 mm, the Dice coefficient increased to 0.548 (95% CI: 0.485–0.612), while the FP patch rate was 0.030 (95% CI: 0.018–0.042), indicating improved performance on clinically relevant lesions. Statistical analysis confirmed that the proposed model was significantly superior to the compared models in both the Dice coefficient and FP patch rate. Although the framework produces patch-level rather than whole-volume segmentations, visual analysis demonstrates accurate localization of aneurysm regions. These results indicate that combining skeleton-guided patch extraction with deep supervision and auxiliary classification improves segmentation performance while providing reliable localization cues for automated aneurysm detection.

Author Biographies

Mario Mata-Castillo, Universidad Autónoma de Ciudad Juárez, Ciudad Juárez, Mexico

Mario Mata is a doctoral candidate at the Autonomous University of Ciudad Juárez (UACJ), Chihuahua, México, and a professor at the Autonomous University of Chihuahua. His research focuses on artificial intelligence in medical imaging, with an emphasis on deep learning-based methods for biomedical image segmentation. His interests span machine learning and the application of AI techniques to improve diagnostic processes and healthcare outcomes. (EMAIL: al221148@alumnos.uacj.mx)

Nelly Gordillo-Castillo, Universidad Autónoma de Ciudad Juárez, Ciudad Juárez, Mexico

Nelly Gordillo-Castillo is a full-time professor at the Autonomous University of Ciudad Juárez (UACJ), México, and Head of the Department of Electrical and Computer Engineering. Her research focuses on medical image processing and artificial intelligence for the detection of neurological pathologies. She is a member of the National System of Researchers (SNII Level 2). (EMAIL: nelly.gordillo@uacj.mx)

Boris Mederos, Universidad Autónoma de Ciudad Juárez, Ciudad Juárez, Mexico

Boris Mederos is a full-time professor at the Autonomous University of Ciudad Juárez. His work is in the area of applied mathematics and artificial intelligence, with interests in machine learning, image processing, pattern recognition, and mathematical modeling. He earned both his Ph.D. and M.Sc. in Mathematics from IMPA, Brazil, and completed postdoctoral appointments at the University of California, Davis, and at IMPA itself. His academic work combines theoretical foundations, computational development, and applications to scientific and technological problems. (EMAIL: boris.mederos@uacj.mx)

Dora-Luz Flores , Universidad Autónoma de Baja California, Mexicali, Mexico

Dora-Luz Flores is a Full Professor and currently serves as the General Coordinator of Research and Graduate Studies at the Universidad Autónoma de Baja California (UABC), Mexico. She has a trajectory in biomedical engineering and research management. Her work focuses on artificial intelligence applied to biological systems, computational modeling, and technology transfer. She has led multiple national and international collaborations, particularly in cross-border initiatives. Her research interests include AI in health, innovation policy, graduate education, and scientific development strategies. (EMAIL: dflores@uabc.edu.mx)

Vicente García, Universidad Autónoma de Ciudad Juárez, Ciudad Juárez, Mexico

Vicente García is a full-time Professor at the Department of Electrical and Computer Engineering at the Universidad Autónoma de Ciudad Juárez (UACJ), Mexico, and Coordinator of the Bachelor’s Program in Data Science and Artificial Intelligence. He holds a Ph.D. degree. His research focuses on Machine Learning and Data Mining, particularly on classification, data preprocessing, data complexity, and predictive model evaluation, with emphasis on class imbalance and bias and fairness in artificial intelligence. (EMAIL: vicente.jimenez@uacj.mx)

José Díaz-Román, Universidad Autónoma de Ciudad Juárez, Ciudad Juárez, Mexico

José David Díaz Román is a full-time professor at the Autonomous University of Ciudad Juárez (UACJ), México, and Coordinator of the Doctorate in Advanced Engineering Sciences. His research interests include digital signal and image processing, artificial intelligence, machine learning, deep learning, biomedical instrumentation, and wireless sensor networks. (EMAIL: david.roman@uacj.mx)

References

[1] M. T. Lawton and G. E. Vates, “Subarachnoid Hemorrhage,” New England Journal of Medicine, vol. 377, no. 3, pp. 257–266, Jul. 2017, doi: 10.1056/NEJMcp1605827.

[2] T. Ingall, K. Asplund, M. Mähönen, and R. Bonita, “A Multinational Comparison of Subarachnoid Hemorrhage Epidemiology in the WHO MONICA Stroke Study,” 2000. [Online]. Available: http://ahajournals.org

[3] N. Etminan and G. J. Rinkel, “Unruptured intracranial aneurysms: Development, rupture and preventive management,” Dec. 01, 2016, Nature Publishing Group. doi: 10.1038/nrneurol.2016.150.

[4] B. M. Howard, R. Hu, J. W. Barrow, and D. L. Barrow, “Comprehensive review of imaging of intracranial aneurysms and angiographically negative subarachnoid hemorrhage,” Neurosurg. Focus, vol. 47, no. 6, pp. 1–13, Dec. 2019, doi: 10.3171/2019.9.FOCUS19653.

[5] Z. Shi et al., “Artificial intelligence in the management of intracranial aneurysms: Current status and future perspectives,” Mar. 01, 2020, American Society of Neuroradiology. doi: 10.3174/AJNR.A6468.

[6] M. Din, S. Agarwal, M. Grzeda, D. A. Wood, M. Modat, and T. C. Booth, “Detection of cerebral aneurysms using artificial intelligence: A systematic review and meta-analysis,” Mar. 01, 2023, BMJ Publishing Group. doi: 10.1136/jnis-2022-019456.

[7] Ž. Bizjak and Ž. Špiclin, “A Systematic Review of Deep-Learning Methods for Intracranial Aneurysm Detection in CT Angiography,” Nov. 01, 2023, Multidisciplinary Digital Publishing Institute (MDPI). doi: 10.3390/biomedicines11112921.

[8] F. Isensee, P. F. Jaeger, S. A. A. Kohl, J. Petersen, and K. H. Maier-Hein, “nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation,” Nat. Methods, vol. 18, no. 2, pp. 203–211, Feb. 2021, doi: 10.1038/s41592-020-01008-z.

[9] J. Chen et al., “TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation.” [Online]. Available: https://github.com/Beckschen/

[10] H. Cao et al., “Swin-Unet: Unet-like Pure Transformer for Medical Image Segmentation,” May 2021, [Online]. Available: http://arxiv.org/abs/2105.05537

[11] K. M. Timmins et al., “Comparing methods of detecting and segmenting unruptured intracranial aneurysms on TOF-MRAS: The ADAM challenge,” Neuroimage, vol. 238, Sep. 2021, doi: 10.1016/j.neuroimage.2021.118216.

[12] S. Ham et al., “Automated detection of intracranial aneurysms using skeleton-based 3D patches, semantic segmentation, and auxiliary classification for overcoming data imbalance in brain TOF-MRA,” Sci. Rep., vol. 13, no. 1, Dec. 2023, doi: 10.1038/s41598-023-38586-9.

[13] Ö. Çiçek, A. Abdulkadir, S. S. Lienkamp, T. Brox, and O. Ronneberger, “3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation,” in Medical Image Computing and Computer-Assisted Intervention – MICCAI 2016: 19th International Conference, Athens, Greece, October 17-21, 2016, Proceedings, Part II, Berlin, Heidelberg: Springer-Verlag, 2016, pp. 424–432. doi: 10.1007/978-3-319-46723-8_49.

[14] A. and G. L. M. and S. J. Cireşan Dan C. and Giusti, “Mitosis Detection in Breast Cancer Histology Images with Deep Neural Networks,” in Medical Image Computing and Computer-Assisted Intervention – MICCAI 2013, I. and S. Y. and B. C. and N. N. Mori Kensaku and Sakuma, Ed., Berlin, Heidelberg: Springer Berlin Heidelberg, 2013, pp. 411–418.

[15] J. N. Kapur, P. K. Sahoo, and A. K. C. Wong, “A new method for gray-level picture thresholding using the entropy of the histogram,” Comput. Vis. Graph. Image Process., vol. 29, no. 3, pp. 273–285, 1985, doi: https://doi.org/10.1016/0734-189X(85)90125-2.

[16] R. M. Haralick, S. R. Sternberg, and X. Zhuang, “Image Analysis Using Mathematical Morphology,” IEEE Trans. Pattern Anal. Mach. Intell., vol. PAMI-9, no. 4, pp. 532–550, 1987, doi: 10.1109/TPAMI.1987.4767941.

[17] P. Maragos and R. Schafer, “Morphological skeleton representation and coding of binary images,” IEEE Trans. Acoust., vol. 34, no. 5, pp. 1228–1244, 1986, doi: 10.1109/TASSP.1986.1164959.

[18] J. Yang, B. Wu, L. Li, P. Cao, and O. Zaiane, “MSDS-UNet: A multi-scale deeply supervised 3D U-Net for automatic segmentation of lung tumor in CT,” Computerized Medical Imaging and Graphics, vol. 92, p. 101957, 2021, doi: https://doi.org/10.1016/j.compmedimag.2021.101957.

[19] C. H. Sudre, W. Li, T. Vercauteren, S. Ourselin, and M. Jorge Cardoso, “Generalised Dice Overlap as a Deep Learning Loss Function for Highly Unbalanced Segmentations,” in Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support, M. J. Cardoso, T. Arbel, G. Carneiro, T. Syeda-Mahmood, J. M. R. S. Tavares, M. Moradi, A. Bradley, H. Greenspan, J. P. Papa, A. Madabhushi, J. C. Nascimento, J. S. Cardoso, V. Belagiannis, and Z. Lu, Eds., Cham: Springer International Publishing, 2017, pp. 240–248.

[20] T.-Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, “Focal Loss for Dense Object Detection,” in 2017 IEEE International Conference on Computer Vision (ICCV), 2017, pp. 2999–3007. doi: 10.1109/ICCV.2017.324.

[21] I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” arXiv preprint arXiv:1711.05101, 2017.

Cover iJOE – Vol 22, No 10 (2026)

Downloads

Published

2026-10-16

How to Cite

Mata-Castillo, M., Gordillo-Castillo, N., Mederos, B., Flores , D.-L., García, V., & Díaz-Román, J. (2026). Deeply Supervised 3D U-Net with Auxiliary Classification for Cerebral Aneurysms Segmentation. International Journal of Online and Biomedical Engineering (iJOE), 22(10), pp. 140–160. https://doi.org/10.3991/ijoe.v22i10.62047

Issue

Vol. 22 No. 10 (2026)

Section

Papers

License

Copyright (c) 2026 Mario Mata-Castillo, Nelly Gordillo-Castillo, Boris Mederos, Dora-Luz Flores , Vicente García, José Díaz-Román

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.