Computational method for identifying and quantifying shape features of human left ventricular remodeling

Abstract

Left ventricular remodeling during the development of heart failure is a strong predictor of cardiovascular mortality. However, methods to objectively quantify remodeling-associated shape changes are not routinely available but may be possible with new computational anatomy tools. In this study, we analyzed and compared multi-detector computed tomographic (MDCT) images of ventricular shape at end-systole (ES) and end-diastole (ED) to determine whether regional structural characteristics could be identified and, as a proof of principle, whether differences in hearts of patients with anterior myocardial infarction (MI) and ischemic cardiomyopathy (ICM) could be distinguished from those with global nonischemic cardiomyopathy (NICM). MDCT images of hearts from 11 patients (5 with ICM) with ejection fractions (EF) < 35% were analyzed. An average ventricular shape model (template) was constructed for each cardiac phase by bringing heart shapes into correspondence using linear and nonlinear image matching algorithms. Next, transformation fields were computed between the template image and individual heart images in the population. Principal component analysis (PCA) method was used to quantify ventricular shape differences described by the transformation vector fields. Statistical analysis of PCA coefficients revealed significant ventricular shape differences at ED (p = 0.03) and ES (p = 0.03). For validation, a second set of 14 EF-matched patients (8 with ICM) were evaluated. The discrimination rule learned from the training data set was able to differentiate ICM from NICM patients (p = 0.008). Application of a novel shape analysis method to in vivo human cardiac images acquired on a clinical scanner is feasible and can quantify regional shape differences at end-systole in remodeled myopathic human myocardium. This approach may be useful in identifying differences in the remodeling process between ICM and NICM populations and possibly in differentiating the populations.

Figures

FIGURE 1

Schematic illustration of the work-flow used for LV shape analysis. After isolating and reconstructing the LV, images were segmented to remove any non-cardiac tissue (not shown). This process improved the accuracy of the image matching algorithm. Next, average LV shape models (templates) were generated separately for ED and ES using affine and non-linear transformations. Once each template was generated, the template was matched to the target LVs using the LDDMM algorithm. Deformation vector fields (the initial momenta) calculated at this step were used to perform statistical shape analysis.

FIGURE 2

(a) Original MDCT image. The image was re-sampled along the superimposed lines to isolate the LV. (b) Examples of re-sampled planes reconstructed from the image depicted in (a).

FIGURE 3

Shape variation model reconstructed for the LV using 11 subjects in the training set at ED (a) and ES (b). In each panel (a and b), top images represent LV shape variation along the combined population largest geometric variation axis (first PC) and bottom images represent LV shape variation along the second largest geometric variation axis (second PC). The middle column in each row represents the reconstructed average template (ED for panel “a” and ES for panel “b”). The first and second columns (from left) in each row represents images synthesized at −2 and −1 σ along the first (top row of each panel) and the second (bottom of each panel) principle component and the fourth and fifth columns (from left) in each row represents images synthesized at +1 and +2 σ along the first (top row of each panel) and the second (bottom of each panel) principle component, respectively. To appreciate the magnitude of variation, the cross-sectional contour of average LV shape has been superimposed on the cross sectional view of all other images (A: anterior, L: lateral, S: septum, and P/I: posterior/inferior).

FIGURE 4

PC analysis of training set initial momenta (a) The scatter plot of coefficients (ICM [star] and NICM [circle]) associated with heart images acquired at ED using the first 3 PCA basis functions (eigenvectors). (b) The scatter plot of coefficients (ICM [star] and NICM [circle]) associated with heart images acquired at ES using the first 3 PCA basis functions (eigenvectors). Each data point represents PC value calculated for one subject. Note that for both cases, the coefficients for the second PC show minimum overlap, and hence can be used to discriminate groups.

FIGURE 5

Surface-normal component map for ED (a) and ES (b) superimposed on the template surface. This map demonstrates the magnitude of outward (warmer colors) and inward movements (cooler colors), as defined by initial velocity vector fields in the direction of the second largest variation subspace. The magnitude and sign of the PC coefficients computed for each group determines the direction of deformations across groups (e.g., negative coefficients reverse the direction of deformation).

FIGURE 6

Reconstructed LV images at ES for ICM (a) and NICM (b) subjects using the second highest geometric variance direction (second PC). The middle column in each row represents the reconstructed image using group mean coefficient for the second PC. The first and second columns (from left) in each row represents images synthesized at −2 and −1 σ of 2nd PC mean coefficient, and the fourth and fifth columns (from left) in each row represents images synthesized at +1 and +2 σ of the 2nd PC mean coefficient, respectively. Subjects in the ICM group, in comparison to the NICM group, exhibit larger within group variation of anterior wall shape (arrows).

FIGURE 7

PC analysis of test set initial momenta. The scatter plot of coefficients associated with test set heart images acquired at ES using the second PC basis function (eigenvectors) derived from the training set. Each data point represents PC value calculated for one subject. The coefficients show good separation (Wilcoxon rank sum p value 5 = 0.008).

FIGURE 8

Map of the voxel-based significance of tissue expansion for NICM group relative to ICM group. The color scale represents the significance of tissue expansion measured in corrected p values, with yellow representing the least significant area.