Motion compensation in the region of the coronary arteries based on partial angle reconstructions from short‐scan CT data. Issue 11 (22nd September 2017)

• Record Type: Journal Article
• Authors: Hahn, Juliane; Bruder, Herbert; Rohkohl, Christopher; Allmendinger, Thomas; Stierstorfer, Karl; Flohr, Thomas; Kachelrieß, Marc
• Is Part Of: Medical physics. Volume 44:Issue 11(2017)
• Page Start: 5795
• Page End: 5813
• Publication Date: 2017-09-22
• Subjects: cardiac CT -- computed tomography (CT) -- motion artifacts -- motion compensation -- motion estimation; Medical physics -- Periodicals; Medical physics; Geneeskunde; Natuurkunde; Toepassingen; Biophysics; Periodicals; Periodicals; Electronic journals; 610.153
• DOI: 10.1002/mp.12514
• Languages: English
• ISSNs: 0094-2405

Abstract

Abstract : Purpose: In order to mitigate motion‐induced artifacts, several motion compensation (MoCo) methods have been developed, which are either able to (a) compensate for severe artifacts, but utilize the data for the reconstruction of several cardiac phases, or (b) improve image quality of a single reconstruction with only moderate motion artifacts. We propose a method combining both benefits: dose efficiency by utilizing only the data needed for a single short‐scan reconstruction while still being able to compensate for severe artifacts. Methods: We introduce a MoCo method, which we call PAMoCo, to improve the visualization of the coronary arteries of a standard coronary CT angiography exam by reducing motion artifacts. As a first step, we segment a region of interest covering a chosen coronary artery. We subdivide a volume covering the whole heart into several stacks, which are sub‐volumes, reconstructed from phase‐correlated short‐scan data acquired during different heart cycles. Motion‐compensated reconstruction is performed for each stack separately, based on partial angle reconstructions, which are derived by dividing the data corresponding to the segmented stack volume into several double‐overlapping sectors. We model motion along the coronary artery center line obtained from segmentation and the temporal dimension by a low‐degree polynomial and create a dense 3D motion vector field (MVF). The parameters defining the MVF are estimated by optimizing an image artifact measuring cost function and we employ a semi‐global optimization routine by re‐initializing the optimization multiple times. The algorithm was evaluated on the basis of a phantom measurement and clinical data. For the phantom measurement an artificial vessel equipped with calcified lesions mounted on a moving robot arm was measured, where typical coronary artery motion patterns for 70 bpm and 90 bpm have been applied. For analysis, we calculated the calcified volume V inside an ROI and measured the maximum vessel diameter d based on cross‐sectional views to compare the performances of standard reconstructions obtained via filtered backprojection (FBP) and PAMoCo reconstructions between 20% and 80% of the cardiac cycle. Further, the new algorithm was applied to six clinical cases of patients with heart rates between 50 bpm and 74 bpm. Standard FBP, PAMoCo reconstructions were performed and compared to best phase FBP reconstructions and another MoCo algorithm, which is based on motion artifact metrics (MAM), via visual inspection. Results: In case of the phantom measurement we found a strong dependence of V and d on the cardiac phase in case of the FBP reconstructions. When applying PAMoCo, V and d became almost constant due to a better discrimination from calcium to vessel and water background and values close to the ground truth have been derived. In the clinical study we chose reconstructions showing strong motion artifacts and observed a substantially improved delineation of the coronary arteries in PAMoCo reconstructions compared to the standard FBP reconstructions and also the MAM images, confirming the findings of the phantom measurement. Conclusions: Due to the fast reconstruction of PAMoCo images and the introduction of a new motion model, we were able to re‐initialize the optimization routine at pre‐selected parameter sets and thereby increase the potential of the MAM algorithm. From the phantom measurement we conclude that PAMoCo performed almost equally well in all cardiac phases and suggest applying the PAMoCo algorithm for single source systems in case of patients with high or irregular heart rates.