Variability of global left ventricular deformation analysis using vendor dependent and independent two-dimensional speckle-tracking software in adults.

BACKGROUND Evaluation of myocardial deformation by two-dimensional speckle-tracking is useful for clinical and research purposes. However, differences may exist among different ultrasound machines, software packages, frame rates, and observers. METHODS Thirty patients underwent echocardiography on both GE (Vivid 9; GE Vingmed Ultrasound AS, Horten, Norway) and Philips (iE33; Philips Ultrasound, Bothell, WA) ultrasound systems. From each study, two sets of images were stored in Digital Imaging and Communications in Medicine format, optimized for strain evaluation: one set of images at the acquisition frame rate (55-90 frames/sec) and one set of images at a compressed frame rate of 30 frames/sec. Vendor-independent software (VIS; TomTec 2D Cardiac Performance Analysis, Munich, Germany) was used to measure strain in multiple directions and was compared with vendor-specific software (GE EchoPAC; GE Vingmed Ultrasound AS). RESULTS Intraobserver and interobserver coefficients of variation ranged from 5.5% to 8.7% for longitudinal strain, from 10.7% to 20.8% for circumferential, and from 15.3% to 33.4% for radial and transverse strain. Strain values obtained using VIS were comparable with those obtained using vendor-specific software for longitudinal strain, regardless of ultrasound machine or frame rate. For circumferential strain, a consistent large bias was observed between VIS and vendor-specific software, with higher values using VIS. Slightly higher strain values were observed by analysis at the acquisition frame rate compared with the low frame rate, but no consistent bias was observed between images from different vendors. CONCLUSIONS Global longitudinal strain consistently showed good reproducibility, while reproducibility was moderate for circumferential strain and poor in the radial direction. Retrospective analysis of legacy Digital Imaging and Communications in Medicine data at 30 frames/sec can be reliably performed for longitudinal strain.