Motion estimation for cardiac functional analysis using two x‐ray computed tomography scans

Abstract

Purpose This work concerns computed tomography (CT)-based cardiac functional analysis (CFA) with a reduced radiation dose. As CT-CFA requires images over the entire heartbeat, the scans are often performed at 10–20% of the tube current settings that are typically used for coronary CT angiography. A large image noise then degrades the accuracy of motion estimation. Moreover, even if the scan was performed during the sinus rhythm, the cardiac motion observed in CT images may not be cyclic with patients with atrial fibrillation. In this study, we propose to use two CT scan data, one for CT angiography at a quiescent phase at a standard dose and the other for CFA over the entire heart beat at a lower dose. Methods We have made the following four modifications to an image-based cardiac motion estimation method we have previously developed for a full-dose retrospectively-gated coronary CT angiography: (1) A full-dose prospectively-gated coronary CT angiography image acquired at the least motion phase was used as the reference image; (2) a three-dimensional median filter was applied to lower-dose retrospectively-gated cardiac images acquired at 20 phases over one heartbeat in order to reduce image noise; (3) the strength of the temporal regularization term was made adaptive; and (4) a one-dimensional temporal filter was applied to the estimated motion vector field in order to decrease jaggy motion patterns. We call the conventional method iME1 and the proposed method iME2 in this paper. Five observers assessed the accuracy of the estimated motion vector field of iME2 and iME1 using a 4-point scale. The observers repeated the assessment with data presented in a new random order one week after the first assessment session. Results The study confirmed that the proposed iME2 was robust against the mismatch of noise levels, contrast enhancement levels, and shapes of the chambers. There was a statistically significant difference between iME2 and iME1 (accuracy score, 2.08±0.81 versus 2.77±0.98, P<0.01) and the improvement by the score of +0.69 seemed clinically relevant. Inter-observer concordance was good: The inter-class correlation coefficient was 0.63 and Kendall’s rank correlation coefficients were in the range of 0.41–0.67 (P<0.01) respectively. Intra-observer reproducibility between sessions was good with the inter-class correlation coefficient of 0.76. Conclusion We have proposed iME2 method for CT-CFA with two CT scans. The observer study verified the robustness and accuracy of iME2 method and its improved performance over iME1 method

Vectors through a cross-sectional image (VCI): A visualization method for four-dimensional motion analysis for cardiac computed tomography

“…Details of the iME algorithm have been reported elsewhere 7,8 and we provide a brief description as follows. First, 3D images at 10 phases were down-sampled from ~512 3 voxels to ~256 3 voxels and one phase (it was the end-diastole at 0% of the R-R interval in this study) was chosen as the reference phase.…”

“…Recently, to evaluate the cardiac wall motion, a new image-based motion-estimation (iME) algorithm has been developed for use with CT. 7,8 Similar to speckle-tracking for 3D echocardiography, iME estimates the regional four-dimensional (4D=3D plus time) motion of the heart from a time series of 3D CT images and outputs motion vector fields. 9,10 One of the challenges in 4D motion analysis is how to display the motion vector field and CT images together in order to capture through-plane motion better.…”

Iterative inversion of deformation vector fields with feedback control

“…Inverse DVFs are needed, together with their respective forward DVFs, to map images, structure contours, or doses back and forth in applications such as 4D image reconstruction, 1 dose accumulation calculations and multimodality treatment planning in adaptive radiotherapy, [2][3][4][5] and cardiac functional analysis. 6 DVF inversion is also a fundamental operation in simultaneous and symmetric registration methods. [7][8][9][10][11] An important consideration is ensuring that the forward and reverse mappings are inverse-consistent.…”

“…We consider numerical inversion of a deformation vector field (DVF). Inverse DVFs are needed, together with their respective forward DVFs, to map images, structure contours, or doses back and forth in applications such as 4D image reconstruction [1], dose accumulation calculations and multi-modality treatment planning in adaptive radiotherapy [2,3,4,5], and cardiac functional analysis [6]. DVF inversion is also a fundamental operation in simultaneous and symmetric registration methods [7,8,9,10,11].…”

“…One of the images may be more adversely affected by noise or artifacts than the other. Such asymmetry makes one-way registration better or worse depending on the mapping direction, due to the high sensitivity of registration methods to noise or variation in imaging conditions [14,2,6]. We consider the asymmetric approach to be an effective means to counterbalancing the asymmetry in image quality.…”