Spatio-temporal nonrigid registration for ultrasound cardiac motion estimation
Summary (3 min read)
Introduction
- The authors evaluated the accuracy of their algorithm using a synthetic sequence generated with an ultrasound simulation package, together with a realistic cardiac motion model.
- The authors work concentrates on two-dimensional (2-D) echocardiography, as it is ubiquitous, and is the most widely used imaging method to assess cardiac function.
- The authors propose using a nonrigid parametric motion estimation algorithm developed to overcome some of the underlying problems inherent to echocardiographic image tracking.
A. Problem Definition
- Let us consider an image sequence with and , where is the intensity at time and position .
- The goal is to find a dense displacement field over the whole sequence; to this end, the authors introduce the deformation function, , which represents the position at time of a point that was at position at time , i.e., the so-called Lagrangian representation.
- In other words, the authors are using the first frame as a spatial reference, implying .
B. Consecutive Registration
- This registration method is described in [37], and is based on the registration of consecutive pairs of images obtained from the sequence, using an algorithm derived from [38].
- This approach calculates the interframe displacement fields .
- The total deformation field, , is then obtained from the contribution of the partial fields.
- Registration is performed twice, in the forward and backward directions, to minimize any error accumulation, and the mean of the two displacements is used as the final result.
- The authors have also imposed a periodicity on the measurements, as the sequence encompasses a complete cycle.
C. Spatio-Temporal Registration
- In contrast to the consecutive registration method, the new algorithm presented in this article works globally on all the images of the sequence simultaneously.
- It searches for a spatio-temporal deformation field, , expressed by a parametric B-spline model.
- The key features of the algorithm are the similarity criterion (Section II-D), and the spatio-temporal deformation model (Section II-F).
- The upper part of Fig. 1 shows three images from the original sequence covering the entire cardiac cycle.
D. Optimization Criterion
- The authors registration procedure seeks a minimum value, , for a criterion, , which is defined as the mean value obtained from the entire sequence of an image similarity criterion, (1) (2) where is the total number of images in the sequence, is the set of coordinates specifying the spatial region of interest, and is the corresponding number of pixels.
- The authors chose to use the SSD criterion because of its simplicity, fast computation time, and smoothness of the resulting criterion space.
- It is necessary to have a continuous version of to be able to calculate the warped sequence, , by interpolation, as well as to be able to evaluate the criterion derivatives.
- The coefficients, , were obtained from the pixel values, , using filtering [39].
- The spline model has the advantage of good accuracy, low computational complexity, and allows for the possibility of evaluating spatial derivatives analytically.
F. Spatio-Temporal Model
- The deformation function, , is represented by a linear model, which is separable in time and space, with parameters, (4) where and define the set of spatial and temporal parameter indices.
- Specifically, the authors used the following basis functions: (5) where and (6) The basis functions, , were placed on a uniform rectangular spatial grid, and the were placed at regularly spaced time intervals.
- These parameters also control the rigidity of the solution.
- In Section III-B2 the authors analyze the influence of the knot spacings in more detail.
G. Motion Field Constraints
- The motion model of (5) can be further constrained by using a priori knowledge of the motion field.
- This increases the robustness of the registration process by taking out superfluous degrees of freedom.
- First, the authors know that the displacement at the reference frame must be zero.
- It depicts the individual basis functions scaled using proper coefficients, as well as the overall trajectory .
H. Multiresolution and Optimization Strategy
- The solution to their registration problem is a deformation field, , that minimizes the criterion, .
- This is found by using a multidimensional optimization algorithm acting on the parameters .
- The projection onto the finer space was achieved using no approximations, thanks to the embedding properties of the underlying B-spline spaces.
- To summarize, the optimization process proceeded in a coarse-to-fine fashion for both the image sequence and the motion field model.
- The convergence speed depends on the number of parameters and the sequence size.
III. EXPERIMENTS WITH SIMULATED DATA
- This section discusses evaluation experiments on simulated data.
- The authors analyzed the benefits of the temporal model and the influence of the different algorithm parameters.
- The use of simulated sequences allows us to quantify the accuracy of the reconstructed motion, which would not be possible to obtain with real data.
- As the true cardiac motion was not available, the authors generated a realistic cardiac motion field.
- The corresponding model was separable, and consisted of two components: an affine spatial component that simulated radial myocardial contraction or expansion, and a temporal component that modulated this movement in a realistic fashion throughout the cardiac cycle.
A. Simulated Sequences
- The authors generated two different sets of simulated sequences using the model mentioned above.
- The authors corrupted the deformed images using different levels of additive Gaussian noise.
- Second, the authors generated a sequence using the FIELD II ultrasound simulation package [43], [44].
- The authors designed the first frame of the scattering map using a real end-diastole image as a template.
- The final image was calculated by summing the responses of all the scatterers, which were specified by their positions and amplitudes [45].
B. Experiments and Results
- This section discusses a series of experiments carried out to evaluate the performance of the algorithm.
- The authors used the same parameter settings for both algorithms: Fig. 6 shows the geometric error for different values of the knot spacing, for a fixed value of .
- The projection error (“ideal” error) decreases with the step size, .
A. Data Description and Methodology
- The authors now describe the use of their algorithm in a clinical setting.
- This process is usually denoted as regional analysis.
- The authors study quantified the function of the basal and mid segments for the inferior (2C view) and septal walls (4C view), a total of 48 segments.
- The authors selected these segments as they were clearly visible in all the sequences.
- The authors also checked that after applying the estimated displacement the segment contours were correctly repositioned in the remaining frames of the sequence, which indicated that the recovered displacement field was consistent with the real motion.
B. Results
- Fig. 9 shows the displacement field at the end of systole (maximum contraction) for a patient with an anterior acute infarct in the apical 2C and 4C views.
- Notice the difference in arrow lengths between the anterior wall (left image, left wall), classified as akinetic, and the basal inferior segment, classified as hypokinetic.
- The change between the normal and the hypokinetic, and between the akinetic and the hypokinetic segments is not so well defined.
- Table II shows the results of the second study that examined the displacement of all segments independently for healthy subjects and patients.
- This effect confirms that the longitudinal displacement decreases from base to apex.
V. DISCUSSION AND CONCLUSION
- The authors presented a new, fully automatic procedure to compute cardiac motion from echocardiographic sequences using nonrigid registration techniques.
- The authors method exploits the temporal coherence of the movement, and estimates the motion field by registering the sequence to a reference frame.
- The key methodological contributions of the present work are as follows.
- Displacement and strain values are consistent with those previously published and obtained with Doppler derived techniques [50]–[52], and Tagged MR data [53], [54].
- This clinical validation should also consider whether the variability in the definition of long axis by the user has any influence on the results.
Did you find this useful? Give us your feedback
Citations
1,434 citations
1,150 citations
Cites background from "Spatio-temporal nonrigid registrati..."
...Gradient based level set algorithms have also been the most popular [14], [19], [21], [61], [62], [64], [65] and only few have used region information [38], [111], [134], [148], [214] or a combination of both [20], [22], [23], [170]....
[...]
689 citations
Cites background from "Spatio-temporal nonrigid registrati..."
...Examples of spatiotemporal image registration of the heart can be found in Grau et al. (2007), Ledesma-Carbayo et al. (2005), Perperidis et al. (2005), Peyrat et al. (2010), and a solution for temporal plantar pressure image sequences registration is presented in Oliveira et al. (2011a)....
[...]
...…(Shekhar et al. 2005), cervical (Staring et al. 2009), heart (Dey et al. 1999; Shekhar and Zagrodsky 2002; Rhode et al. 2003; Shekhar et al. 2004; Ledesma-Carbayo et al. 2005; Grau et al. 2007; Huang et al. 2009), pelvis (Hamilton et al. 1999; Shen 2004, 2007), wrist (Giessen et al. 2009),…...
[...]
..., 2009), heart (Dey et al., 1999; Shekhar and Zagrodsky, 2002; Rhode et al., 2003; Shekhar et al., 2004; Ledesma-Carbayo et al., 2005; Grau et al., 2007; Huang et al., 2009), pelvis (Hamilton et al....
[...]
183 citations
167 citations
Cites methods from "Spatio-temporal nonrigid registrati..."
...On the other hand, different image registration approaches have been applied to B-mode image sequences, such as the method of optical flow [11]–[13] or elastic registration between subsequent image frames [14]....
[...]
References
8,255 citations
5,490 citations
Additional excerpts
...As shown in [11], [38], and [40], B-splines constitute a good choice for the spatial basis functions, ....
[...]
2,801 citations
"Spatio-temporal nonrigid registrati..." refers methods in this paper
...We measured the accuracy of the motion estimation using a warping index [46], which was defined as the mean geometric error in pixels between the true and the recovered deformation, defined as , where is the region of interest, the myocardium in our case, is...
[...]
2,340 citations
"Spatio-temporal nonrigid registrati..." refers methods in this paper
...Second, we generated a sequence using the FIELD II ultrasound simulation package [43], [44]....
[...]
1,976 citations
"Spatio-temporal nonrigid registrati..." refers methods in this paper
...Second, we generated a sequence using the FIELD II ultrasound simulation package [ 43 ], [44]....
[...]
Related Papers (5)
Frequently Asked Questions (9)
Q2. What have the authors stated for future works in "Spatio-temporal nonrigid registration for ultrasound cardiac motion estimation" ?
This preliminary clinical evaluation encourages further research on the use of the proposed method to derive quantitative parameters to indicate the presence of ischemic disease. Potential applications include motion estimation from other cardiac imaging modalities, cardiac sequence segmentation guided by registration, and coding/compression of movement components.
Q3. What is the effect of noise on the optimum values?
When noise is present, the optimum values should be chosen as a compromise between the approximation error, which is dominant for coarser grids, and the lack of regularization for fine grids, which increases the effect of noise.
Q4. Why did the authors use B-splines for the spatial basis functions?
The authors also used B-splines for the temporal basis functions, , [4], [11], [16], [17], because of their computational simplicity, good approximation properties, and implicit smoothness (minimum curvature property).
Q5. What is the advantage of using B-splines for the spatial basis functions?
The authors found that temporal B-splines performed at least as well as harmonic functions (as used in [1], [2], and [5]) in terms of registration accuracy, with the advantage that the criterion minimization was easier, thanks to their compact support [41].
Q6. Why did the authors not obtain enough data to perform a meaningful statistical study?
because of the small number of cases and the large number of categories, the authors did not obtain enough data to perform a meaningful statistical study.
Q7. What is the method used to estimate the motion field?
Their method exploits the temporal coherence of the movement, and estimates the motion field by registering the sequence to a reference frame.
Q8. What is the way to achieve the projection onto the finer space?
The projection onto the finer space was achieved using no approximations, thanks to the embedding properties of the underlying B-spline spaces.
Q9. What did the authors find that was not worth the additional computational effort?
The authors found that, using cubic B-splines in the spatial direction did not improve the accuracy significantly and was not worth the additional computational effort.