Segmentation of the left ventricular endocardium from magnetic resonance images by using different statistical shape models

Abstract

We evaluate in this paper different strategies for the construction of a statistical shape model (SSM) of the left ventricle (LV) to be used for segmentation in cardiac magnetic resonance (CMR) images. From a large database of LV surfaces obtained throughout the cardiac cycle from 3D echocardiographic (3DE) LV images, different LV shape models were built by varying the considered phase in the cardiac cycle and the registration procedure employed for surface alignment. Principal component analysis was computed to describe the statistical variability of the SSMs, which were then deformed by applying an active shape model (ASM) approach to segment the LV endocardium in CMR images of 45 patients. Segmentation performance was evaluated by comparing LV volumes derived by ASM segmentation with different SSMs and those obtained by manual tracing, considered as a reference. A high correlation (r(2)>0.92) was found in all cases, with better results when using the SSM models comprising more than one frame of the cardiac cycle.

A statistical shape model of the left ventricle from real-time 3D echocardiography and its application to myocardial segmentation of cardiac magnetic resonance images

“…Our study confirmed that 3DE represents a convenient choice for the training database, being noninvasive, widely used in clinics and allowing the 3D spatio-temporal representation of the cardiac chambers, as also previously suggested [27,32]. This strategy potentially overcomes the limitation of creating 3D models from interpolated 2D contours from CMR images, which in standard acquisitions are characterized by submillimetric in-plane resolution but slice thickness up to 8 mm.…”

“…The choice of considering multiple frames to construct the LV PDM is supported by a previous research [32], where we evaluated the effect of different registration strategies to build the LV PDM and found optimal performances when the PDM included more than one frame within the cardiac cycle.…”

Automatic initialization and quality control of large-scale cardiac MRI segmentations

“…To enable automatic initialization, an accurate identification of key anatomical landmarks is an essential step in cardiac MRI segmentation, as these points will estimate the initial pose, scale and rotation of the model. Generally, as shown in Table 2, relevant landmark points include the mitral valve hinge points (defining the base line position), the apex and the central axis position, which are often used to initialize the segmentation process (Tobon-Gomez et al, 2012;Tsadok et al, 2013;Piazzese et al, 2016). These points are usually defined manually, which is impractical for large population studies.…”

3D right ventricular endocardium segmentation in cardiac magnetic resonance images by using a new inter-modality statistical shape modelling method

“…Recently, we developed an inter-modality statistical shape modelling (SSM) approach for segmenting the LV. Briefly, 3D surfaces, obtained from real-time threedimensional echocardiographic (3DE) data, were used to train the SM [19]. The LV cavity was then detected by applying the SM in CMR short-axis (SAX) and long-axis (LAX) scans [20,21].…”