作者: Mathieu De Craene , Gemma Piella , Oscar Camara , Nicolas Duchateau , Etelvino Silva
DOI: 10.1016/J.MEDIA.2011.10.006
关键词: Computer vision 、 Algorithm 、 Displacement (vector) 、 Infinitesimal strain theory 、 Mathematics 、 Free-form deformation 、 Displacement field 、 Artificial intelligence 、 Similarity (geometry) 、 Energy functional 、 Metric (mathematics) 、 Temporal resolution
摘要: This paper presents a new registration algorithm, called Temporal Diffeomorphic Free Form Deformation (TDFFD), and its application to motion strain quantification from sequence of 3D ultrasound (US) images. The originality our approach resides in enforcing time consistency by representing the 4D velocity field as sum continuous spatiotemporal B-Spline kernels. displacement is then recovered through forward Eulerian integration non-stationary field. tensor computed locally using spatial derivatives reconstructed energy functional considered this weighs two terms: image similarity regularization term. metric squared differences between intensities each frame reference one. Any can be chosen reference. term based on incompressibility myocardial tissue. TDFFD was compared pairwise FFD 3D+t FFD, both fields, set synthetic US images with different noise levels. showed increased robustness these state-of-the-art algorithms. also proved more resistant reduced temporal resolution when decimating sequence. Finally, dataset used determine optimal settings algorithm. Subsequently, applied database cardiac left ventricle acquired 9 healthy volunteers 13 patients treated Cardiac Resynchronization Therapy (CRT). On cases, uniform patterns were observed over all segments, physiologically expected. CRT patients, improvement synchrony regional longitudinal correlated clinical outcome quantified reduction end-systolic ventricular volume at follow-up (6 12months), showing potential proposed algorithm for assessment CRT.