Motion compensation for cone-beam CT using Fourier consistency conditions. (18th August 2017)

• Record Type: Journal Article
• Authors: Berger, M; Xia, Y; Aichinger, W; Mentl, K; Unberath, M; Aichert, A; Riess, C; Hornegger, J; Fahrig, R; Maier, A
• Is Part Of: Physics in medicine & biology. Volume 62:Number 17(2017:Sep.)
• Page Start: 7181
• Page End: 7215
• Publication Date: 2017-08-18
• Subjects: Fourier consistency conditions -- cone-beam CT -- motion estimation -- motion correction; Biophysics -- Periodicals; Medical physics -- Periodicals; 610.153
• DOI: 10.1088/1361-6560/aa8129
• Languages: English
• ISSNs: 0031-9155

Abstract

Abstract: In cone-beam CT, involuntary patient motion and inaccurate or irreproducible scanner motion substantially degrades image quality. To avoid artifacts this motion needs to be estimated and compensated during image reconstruction. In previous work we showed that Fourier consistency conditions (FCC) can be used in fan-beam CT to estimate motion in the sinogram domain. This work extends the FCC to 3 D cone-beam CT. We derive an efficient cost function to compensate for 3 D motion using 2 D detector translations. The extended FCC method have been tested with five translational motion patterns, using a challenging numerical phantom. We evaluated the root-mean-square-error and the structural-similarity-index between motion corrected and motion-free reconstructions. Additionally, we computed the mean-absolute-difference (MAD) between the estimated and the ground-truth motion. The practical applicability of the method is demonstrated by application to respiratory motion estimation in rotational angiography, but also to motion correction for weight-bearing imaging of knees. Where the latter makes use of a specifically modified FCC version which is robust to axial truncation. The results show a great reduction of motion artifacts. Accurate estimation results were achieved with a maximum MAD value of 708 μ m and 1184 μ m for motion along the vertical and horizontal detector direction, respectively. The image quality of reconstructions obtained with the proposed method is close to that of motion corrected reconstructions based on the ground-truth motion. Simulations using noise-free and noisy data demonstrate that FCC are robust to noise. Even high-frequency motion was accurately estimated leading to a considerable reduction of streaking artifacts. The method is purely image-based and therefore independent of any auxiliary data.