Local Phase-Based Fast Ray Features for Automatic Left Ventricle Apical View Detection in 3D Echocardiography

• Published in: Medical Computer Vision. Large Data in Medical Imaging pp. 119 - 129
• Main Authors: Domingos, João S., Lima, Eduardo, Leeson, Paul, Noble, J. Alison
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing 2014
• Series: Lecture Notes in Computer Science
• Subjects: Detector Views; Monogenic Signal; Nominal Slice; Ray Features; Standard Apical Views
• ISBN: 9783319055299; 3319055291
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-05530-5_12

3D echocardiography is an imaging modality that enables a more complete and rapid cardiac function assessment. However, as a time-consuming procedure, it calls upon automatic view detection to enable fast 3D volume navigation and analysis. We propose a combinatorial model- and machine learning-based left ventricle (LV) apical view detection method consisting of three steps: first, multiscale local phase-based 3D boundary detection is used to fit a deformable model to the boundaries of the LV blood pool. After candidate slice extraction around the derived mid axis of the LV segmentation, we propose the use of local phase-based Fast Ray features to complement conventional Haar features in an AdaBoost-based framework for automated standardized LV apical view detection. Evaluation performed on a combination of healthy volunteers and clinical patients with different image quality and ultrasound probes show that apical plane views can be accurately identified in a 360 degree swipe of 3D frames.