The optimization of parameters and matching point pairs in the 3D reconstruction of coronary artery. (May 2021)

• Record Type: Journal Article
• Authors: Tong, Jijun; Wang, Fangliang; Li, Meng; Xia, Shudong; Lin, Wenting
• Is Part Of: Biomedical signal processing and control. Volume 67(2021)
• Publication Date: 2021-05
• Subjects: 3D reconstruction of coronary artery -- Parameter-adjusting LM algorithm -- Dynamic programming -- Reconstruction optimization; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.102534
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A parameter-adjusting Levenberg-Marquarelt(PALM) algorithm combined with a trust region method is proposed to optimize the geometric transformation matrix. Three inherent problems in conjugate point pair matching with epipolar constraints are listed. Solving globally optimal conjugate point pair matching sequence by dynamic programming. Abstract: In coronary angiography, the accuracy of the geometric transformation matrix and the matching of corresponding points between two angiographic images with different angles is the key to achieve 3D reconstruction of coronary. However, due to the movement of heart and hospital bed which commonly occurs during clinical practice, the geometric transformation matrix obtained directly from the imaging system can not well reflect the spatial geometric relationship between different imaging coordinate systems. Therefore, according to the principle of minimizing the error between the actual vessel and projection of the 3D reconstructed vessel, this paper proposes a parameter-adjusting Levenberg-Marquarelt(PALM) algorithm combined with a trust region method to optimize the geometric transformation matrix. As for the matching of corresponding points, the traditional method is achieved by epipolar constraint, which easily leads to wrong matching point pairs. Therefore, this paper further adopts the error matrix defined by the epipolar matching error to achieve global optimal matching point pairs of vascular through dynamic programming while considering the constraints of the vascular smoothness. The experiments show that the 3D reconstruction accuracy has been dramatically improved in terms of mean back projection errors, which confirms the validity and applicability of the algorithm put forward in this paper.