Direct estimation of global longitudinal strain from echocardiograms using a logarithm-scaled Fourier magnitude correlation

We present a new algorithm, the logarithmic-transform correlation (LTC) method, for measuring global longitudinal strain (GLS) and global longitudinal strain rate (GLSr) from 2D echocardiograms. In contrast to conventional GLS computation methods, our approach does not require boundary segmentation and regularization. The algorithm was benchmarked against conventional GLS methods using synthetic left ventricle (LV) ultrasound images generated to represent five different ischemia conditions each on five different ultrasound imaging systems. LV GLS and GLSr measurement error was assessed as a function of contrast-to-noise (CNR) ratio using mean absolute error (MAE) and root-mean-square error (RMSE). Our algorithm showed strong agreement to the ground truth for both GLS ($R^2$=0.91) and GLSr ($R^2$=0.85). Conventional GLS algorithms showed less agreement for GLS ($R^2$=0.7) and GLSr ($R^2$=0.7). In addition, our method was unaffected by CNR. A 200% improvement in both MAE and RMSE was observed compared to conventional GLS algorithms. A clinical demonstration was performed using a 54-subject cohort of pediatric patients (20 subjects with cardiomyopathy, 34 controls). Our method distinguished between normal and abnormal left ventricular function with an AUC = 0.85, a 10% improvement over conventional GLS algorithms.