An automated pipeline for regional cardiac strain estimation from volumetric ultrasound data

Strain estimation techniques can provide valuable insights into regional cardiac (dys)function. However, a manual delineation of the left ventricle is required, intrinsically leading to a time consuming analysis with significant observer variability. In this study we therefore propose a fully automated strain estimation strategy with minimal user input, and illustrate its applicability in a realistic clinical setting. To test the proposed approach, volumetric data (32±3 Hz) was recorded in 3 healthy volunteers and 3 patients with acute myocardial infarction. User interaction time was considerably shorter than manual contouring (19±3s vs 390±90s) and eliminated observer variability. The fully automated strain estimation pipeline showed strain estimates in both subject groups that were in agreement with conventional analysis. Comparing the infarct regions against the remote regions in the patient group, end-systolic strain was reduced by 13.7%, 3.8% and 8.5% for the automated method, and by 12.3%, 5.0% and 9.0% if manually contoured, for the radial, longitudinal and circumferential strain component respectively. Furthermore, the strain curves of both investigated strategies matched well. The proposed automated approach is an attractive solution for clinical practice.