Combined tag tracking and strain reconstruction from tagged cardiac MR images without user‐defined myocardial contours

Abstract

To develop an unsupervised method for measuring quantitative three-dimensional regional strain in the left ventricular wall from tagged cardiac MR images. A total of 10 normal human volunteers and eight patients with myocardial infarction were imaged using a parallel tagged imaging protocol. Each study was analyzed using the combined tag tracking and strain reconstruction (COTTER) algorithm. In contrast to existing techniques, which first track tag lines independently in each slice, then reconstruct myocardial deformation, the COTTER algorithm fits a three-dimensional cardiac deformation model directly to the image data. This approach ensures that tag line positions identified in the image data are consistent from slice to slice. A total of 10 imaging studies (six normals, four patients) were used to optimize parameters of the COTTER algorithm. In the remaining eight imaging studies, the root-mean-square difference between tags tracked by COTTER and user-supervised analysis was 0.66 pixels at end-systole. The correlation coefficient between circumferential shortening strains at end-systole computed by COTTER and user-supervised analysis was 0.84 (P < 0.005) at the midwall. The COTTER algorithm can compute accurate measurements of three-dimensional regional strain without user supervision.