Towards online real-time strain estimation in volumetric us data: Feasibility study and initial clinical validation

Strain-based functional indices have been shown to provide superior performance in assessing global cardiac function when compared with classical volume-based metrics, such as ejection fraction. As in clinical practice global strain is typically used as an index of overall cardiac performance, local tracking algorithms, which are typically computationally intensive, could be substituted by more global approaches. We therefore propose to take advantage of a fast tracking method which uses an optical-flow algorithm on an anatomical ROI to estimate the global cardiac (affine) motion between consecutive frames. The proposed approach was tested in 19 RT3DE exams by assessing the global area strain (GAS), which combines both longitudinal and circumferential deformation. The agreement between the automatic tracking results and the reference show moderate correlation (r=0.698), while Bland-Altman analysis ([μ±1.96σ]=6.0±9.29%) revealed a significant bias, although having competitive limits of agreement when compared with the inter-observer variability ([μ±1.96σ]=0.85±9.52%). The proposed approach takes less than 1 second to perform the tracking between two subsequent frames, in a MATLAB implementation. These preliminary results point towards the feasibility of online estimation of global deformation parameters without any user intervention and near real-time.