Centerline Extraction and 3D Reconstruction of Coronary Angiography Based on Attention 3D U-Net

DOI: https://doi.org/10.62517/jmhs.202505415

Author(s)

Yushuo Mu1, Mingming Gong2,*

Affiliation(s)

1SArtificial Intelligence and Software Engineering, Henan University of Technology, Zhengzhou, Henan, China 2iFLYTEK Co. Ltd., Hefei, Anhui, China *Corresponding Author

Abstract

Coronary artery disease ranks among the leading causes of death globally, and the three-dimensional morphology of atherosclerotic plaques is a critical factor in assessing their rupture risk. To achieve precise 3D modeling of coronary arteries and plaques, this paper proposes an automated reconstruction method integrating angiographic and OCT (Optical Coherence Tomography) images. First, an Attention 3D U-Net network is employed to segment vessels and extract centerlines from angiographic images; Subsequently, based on dual-view projections from the XOZ and YOZ planes, a smooth and continuous three-dimensional centerline is reconstructed through parametric curve fitting and nonlinear optimization techniques. Finally, using this centerline as a spatial reference, OCT cross-sectional sequences are aligned, and the Marching Cubes algorithm is employed to generate a three-dimensional model of the coronary artery, incorporating distinct plaque components. Experimental results demonstrate that this method effectively restores three-dimensional vascular structures from two-dimensional projections. The reconstructed models exhibit excellent geometric plausibility and continuity, accurately reflecting actual anatomical morphology. This study provides a reliable three-dimensional geometric foundation for precise quantitative analysis of coronary lesions and biomechanical assessment of plaque rupture risk.

Keywords

Deep Learning; 3D Reconstruction; Coronary Artery; Attention 3D U-Net; Marching Cubes Algorithm

References

[1]Covani M, Niccoli G ,Fujimoto D , et al. Plaque Vulnerability and Cardiovascular Risk Factor Burden in Acute Coronary Syndrome: An Optical Coherence Tomography Analysis. Journal of the American College of Cardiology, 2025, 86 (2): 77-89. [2]Castaldi G, Zormpas G, Frederiks P, et al. The Rise of Optical Coherent Tomography in Intracoronary Imaging: An Overview of Current Technology, Limitations, and Future Perspectives. Reviews in cardiovascular medicine, 2025, 26 (8): 38123. [3]Xu Q, Ma Z C, He N, et al. DCSAU-Net: A deeper and more compact split-attention U-Net for medical image segmentation. Computers in biology and medicine, 2023, 154 106626-106626. [4]Amandeep K, Guanfang D. A Complete Review on Image Denoising Techniques for Medical Images. Neural Processing Letters, 2023, 55 (6): 7807-7850. [5]Ronneberger O, Fischer P, Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation. CoRR, 2015, abs/1505.04597 [6]Ilesanmi E A, Ilesanmi O T, Ajayi O B. Reviewing 3D convolutional neural network approaches for medical image segmentation. Heliyon, 2024, 10 (6): e27398-. [7]Jijun T, Shuai X, Fangliang W, et al. 3D Reconstruction with Coronary Artery Based on Curve Descriptor and Projection Geometry-Constrained Vasculature Matching. Information, 2022, 13 (1): 38-38. [8]Materka A, Jurek J. Using Deep Learning and B-Splines to Model Blood Vessel Lumen from 3D Images. Sensors (Basel, Switzerland), 2024, 24 (3) [9]Gu L S ,Adhinarta K J ,Bessmeltsev M , et al. Frenet-Serret Frame-based Decomposition for Part Segmentation of 3D Curvilinear Structures. IEEE transactions on medical imaging, 2025, PP [10]Mittal R, Malik V, Singla G, et al. 3D reconstruction of brain tumors from 2D MRI scans: An improved marching cube algorithm. Biomedical Signal Processing and Control, 2024, 91 105901.