Article Details

Abstract

In order to solve the difficulties in aneurysm segmentation such as complex geometric structures and fuzzy boundaries, a multi-module fusion network (CREANet) is proposed. The network employs a curvature-based spectral analysis point recombination strategy to enhance the extraction ability of geometric features. Then, a multi-level retrospective feature optimization module is used to preserve multi-scale details and facilitate the flow of multi-dimensional spatial information. An edge perception enhancement module is introduced to enhance the boundary recognition ability by explicitly predicting edge probabilities. Results on the IntrA dataset show that the network achieves an intersection over union of 95.06% in vessel segmentation and 84.42% in aneurysm segmentation, with a classification accuracy of 94.56%, outperforming several state-of-the-art models. In addition, on ModelNet40 and ShapeNetPart datasets, the overall classification accuracy of the method is 93.4%, and the average intersection over union is 84.7%, which proves the robustness and versatility of the method. This study provides an effective scheme for the accurate segmentation of aneurysms, which has the potential for clinical application.

Keywords

intracranial aneurysm; Deep Learning; curvature traversal; retrospective optimization; boundary enhancement; point cloud

References

1. [1] Cheng M Q，Xiao N，Yuan H，et al. Automatic Intracranial Aneurysm Segmentation Based on Spatial Information Fusion Feature from 3D-RA Using U-Net [C]//2021 IEEE International Conference on Mechatronics and Automation (ICMA)，2021：236-241.
2. [2] Li B H，Guo J Z，Liu H S，et al. UniScene: Unified Occupancy-centric Driving Scene Generation [C]//2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR)，2025：11971-11981.
3. [3] Yu J H，Zhang C Y，Wang H，et al. 3D Medical Point Transformer：Introducing Convolution to Attention Networks for Medical Point Cloud Analysis [J/OL].arXiv:2112.04863 [eess.IV].https://arxiv.org/abs/2112.04863.
4. [4] Cao R F，Zhang D W，Wei P J，et al. PMMNet: A Dual Branch Fusion Network of Point Cloud and Multi-view for Intracranial Aneurysm Classification and Segmentation [J].IEEE Journal of Biomedical and Health Informatics，2025，29（5）：3137-3147.
5. [5] Chen T X，Zhou X D，Tan Z T，et al. Zig-RiR: Zigzag RWKV-in-RWKV for Efficient Medical Image Segmentation [J].IEEE Transactions on Medical Imaging，2025，44（8）：3245-3257.
6. [6] Zhang W Y，Lu F X，Su H J，et al. Dual-branch Multi-information Aggregation Network with Transformer and Convolution for Polyp Segmentation [J/OL].Computers in Biology and Medicine，2024，168：107760.https://doi.org/10.1016/j.compbiomed.2023.107760.
7. [7] He Y F，Guo P F，Tang Y C，et al. VISTA3D: A Unified Segmentation Foundation Model for 3D Medical Imaging [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR)，2025：20863-20873.
8. [8] Perera S，Navard P，Yilmaz A. SegFormer3D: An Efficient Transformer for 3D Medical Image Segmentation [C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)，2024：4981-4988.
9. [9] Lai X，Liu J H，Jiang L，et al. Stratified Transformer for 3D Point Cloud Segmentation [C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR)，2022：8490-8499.
10. [10] Wu X Y，Jiang L，Wang P-S，et al. Point Transformer V3: Simpler, Faster, Stronger [C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR)，2024：4840-4851.
11. [11] Shao D，Lu X Q，Liu X. 3D Intracranial Aneurysm Classification and Segmentation Via Unsupervised Dual-Branch Learning [J].IEEE Journal of Biomedical and Health Informatics，2023，27（4）：1770-1779.
12. [12] Liu Y F，Li W Y，Liu J，et al. GRAB-Net: Graph-based Boundary-aware Network for Medical Point Cloud Segmentation [J].IEEE Transactions on Medical Imaging，2023，42（9）：2776-2786.
13. [13] Guo M-H，Cai J-X，Liu Z-N，et al. PCT: Point Cloud Transformer [J].Computational Visual Media，2021，7（2）：187-199.
14. [14] Zhang Y H，Wang Z H，Han J Ll，et al. GS-Net：Generalizable Plug-and-play 3D Gaussian Splatting Module [J/OL].arXiv:2409.11307 [cs.CV].https://arxiv.org/abs/2409.11307.
15. [15] Cao H Z，Ma Y X，Zhan R H，et al. Adaptive Surface Fitting Convolution for 3D Point Cloud Analysis [C]//2020 5th International Conference on Automation，Control and Robotics Engineering (CACRE)，2020：732-738.
16. [16] Zeng Z Y，Dong M Y，Zhou J，et al. DeepLA-Net: Very Deep Local Aggregation Networks for Point Cloud Analysis [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)，2025：1330-1341.
17. [17] Xu M T，Ding R Y，Zhao H S，et al. PAConv: Position Adaptive Convolution with Dynamic kernel assembling on Point Clouds [C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR)，2021：3172-3181.
18. [18] Engel N，Belagiannis V，Dietmayer K. Point Transformer [J].IEEE Access，2021，9：134826-134840.
19. [19] Zhang C，Wan H C，Shen X Y，et al. PatchFormer: An Efficient Point Transformer with Patch Attention [C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR)，2022：11789-11798.
20. [20] Qian G C，Li Y C，Peng H W，et al. PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategie [J/OL].arXiv:2206.04670 [cs.CV].https://arxiv.org/abs/2206.04670.

About the authors

Wang Yu (2004-), female, Han ethnicity, from Poyang, Jiangxi Province, undergraduate studies, research direction: medical image segmentation; Corresponding author: Guo Lilin (1990-), female, Han ethnicity, from Xinzhou, Shanxi Province, lecturer, Doctor of Engineering, research direction: computer vision, intelligent medical image analysis, multimodal assisted diagnosis and prediction.