Model‐based registration for dynamic cardiac perfusion MRI

Abstract

Purpose:To assess the accuracy of a model-based approach for registration of myocardial dynamic contrastenhanced (DCE)-MRI corrupted by respiratory motion. Materials and Methods:Ten patients were scanned for myocardial perfusion on 3T or 1.5T scanners, and short-and long-axis slices were acquired. Interframe registration was done using an iterative model-based method in conjunction with a mean square difference metric. The method was tested by comparing the absolute motion before and after registration, as determined from manually registered images. Regional flow indices of myocardium calculated from the manually registered data were compared with those obtained with the model-based registration technique. Results:The mean absolute motion of the heart for the short-axis data sets over all the time frames decreased from 5.3 Ϯ 5.2 mm (3.3 Ϯ 3.1 pixels) to 0.8 Ϯ 1.3 mm (0.5 Ϯ 0.7 pixels) in the vertical direction, and from 3.0 Ϯ 3.7 mm (1.7 Ϯ 2.1 pixels) to 0.9 Ϯ 1.2 mm (0.5 Ϯ 0.7 pixels) in the horizontal direction. A mean absolute improvement of 77% over all the data sets was observed in the estimation of the regional perfusion flow indices of the tissue as compared to those obtained from manual registration. Similar results were obtained with two-chamber-view long-axis data sets. Conclusion:The model-based registration method for DCE cardiac data is comparable to manual registration and offers a unique registration method that reduces errors in the quantification of myocardial perfusion parameters as compared to those obtained from manual registration. DYNAMIC MR EVALUATION of myocardial perfusion is becoming a powerful tool for the detection of coronary artery disease (1-3). The standard approach is to rapidly acquire T1-weighted images to track the uptake and washout of the contrast agent Gd-DTPA. Multiple 2D images are acquired each heartbeat. The slices are acquired using ECG-gated sequences so that a given slice is always acquired during the same phase of the cardiac cycle. The time-series data are analyzed visually or with semiquantitative or quantitative models (4,5). Depending on the analysis technique used, at least 20 -40 seconds of data are needed to determine regional perfusion values in the left ventricular (LV) myocardium. Many patients cannot hold their breath long enough to provide a motion-free study. A breathhold of any appreciable length becomes more difficult when pharmacological stress is induced by vasodilating agents, which is necessary for detection of coronary artery disease. Respiration causes motion of the heart, which makes qualitative visual analysis difficult and can cause incorrect estimation of semiquantitative and quantitative parameters.A number of registration methods have been proposed to correct the motion of the heart in dynamic contrast MRI acquisitions (6 -10). In the method of Bidaut and Vallee (6), the mean squared difference between images in the perfusion sequence and a reference image in a region defined by a cardiac mask is minimized. The initial reference image is c...

Segmentation-Driven Image Registration-Application to 4D DCE-MRI Recordings of the Moving Kidneys

“…(iv) Sum-of-squared differences (SSD) of a compartment model to the signal intensity curves within the kidney. The latter two criteria have been used in several related publications for evaluation of co-registration of DCE-MRI image time series, and they are in this respect considered as a gold standard for peformance evaluation [2], [22], [30], [33], [48]. The temporal smoothness (iii) as an evaluation criteria is based Fig.…”

“…In this way, GFR estimates can become strongly biased or even invalidated [55]. To perform motion correction, affine registration has been used by several authors [2], [9]. Affine registration can also be used as an initialization step to a supplementary deformable registration.…”

“…We claim to represent the first work exploring the Mahalanobis distance and a training set for a 4D joint registration and segmentation, in particular for artifact and noise loaded DCE-MRI images. Thus, this work represents a major step towards an integrated model with combined registration, segmentation and compartment modeling, and can further be extended towards model-driven deformable registration in DCE-MRI processing [2], [9], [22].…”

“…Some studies, however, have reported a registration procedure that incorporates a temporal smoothness constraint on the obtained time curves. This has been accomplished in terms of a least squares fitting of the recorded time courses to a parameterized solution of a pharmacokinetic multi-compartment model [2], [9], [22]. Our model is related to those studies by introducing another data term into the registration, sensitive to temporal dynamics.…”

Evaluation of Rigid and Non-rigid Motion Compensation of Cardiac Perfusion MRI

“…The majority of them focuses to compensate the rigid-body motion using image registration [4][5][6][7][8][9][10], while an exception is recently published [11] with the ICA (independent component analysis) being used to analyze the myocardium intensity patterns and to estimate the 2D translation. To correct the possible non-rigid deformation of myocardium during the contrast uptake, methods based on active contour [12] and active shape model [13] have been developed.…”