投影(关系代数)
计算机视觉
计算机科学
人工智能
网(多面体)
端到端原则
计算机图形学(图像)
数学
算法
几何学
作者
Xuan Zhou,Yuedong Liu,Cunfeng Wei
出处
期刊:
日期:2025-01-30
卷期号:30 (2)
摘要
Cone-beam CT (CBCT) is commonly used in industry due to the advantage of large field of view and high imaging efficiency. Robot CT is a unique CBCT system which consists of two independent manipulators carrying the source and detector. It has the ability to scan in arbitraty trajectory while it is hard to gain full-angle projection data. The Parker weighting is usually ultilized in the reconstruction of short scan data, while short scan reconstruction under CB geometry will introduce novel artifacts. In this study, we first analyze the manifestation of mentioned short scan cone beam artifacts, then we develop a projection completion method based on generation of re-projection images with initial reconstruction. After that, then self-optimization of scanned data is adopted to eliminate the artifacts existed in re-projection data through the employment of a residual U-net architecture. Last, we evaluate the performance of proposed method on both simulation and experiment dataset. The experimental results show that this scheme can improve reconstruction quality by about 9% compared to the short scan reconstruction results.
科研通智能强力驱动
Strongly Powered by AbleSci AI