Four-Chamber Heart Modeling and Automatic Segmentation for 3-D Cardiac CT Volumes Using Marginal Space Learning and Steerable Features

Abstract

We propose an automatic four-chamber heart segmentation system for the quantitative functional analysis of the heart from cardiac computed tomography (CT) volumes. Two topics are discussed: heart modeling and automatic model fitting to an unseen volume. Heart modeling is a nontrivial task since the heart is a complex nonrigid organ. The model must be anatomically accurate, allow manual editing, and provide sufficient information to guide automatic detection and segmentation. Unlike previous work, we explicitly represent important landmarks (such as the valves and the ventricular septum cusps) among the control points of the model. The control points can be detected reliably to guide the automatic model fitting process. Using this model, we develop an efficient and robust approach for automatic heart chamber segmentation in 3-D CT volumes. We formulate the segmentation as a two-step learning problem: anatomical structure localization and boundary delineation. In both steps, we exploit the recent advances in learning discriminative models. A novel algorithm, marginal space learning (MSL), is introduced to solve the 9-D similarity transformation search problem for localizing the heart chambers. After determining the pose of the heart chambers, we estimate the 3-D shape through learning-based boundary delineation. The proposed method has been extensively tested on the largest dataset (with 323 volumes from 137 patients) ever reported in the literature. To the best of our knowledge, our system is the fastest with a speed of 4.0 s per volume (on a dual-core 3.2-GHz processor) for the automatic segmentation of all four chambers.

Deep learning analysis of the myocardium in coronary CT angiography for identification of patients with functionally significant coronary artery stenosis

“…They can be divided in boundary-based (e.g. Marie-Pierre (2006); Zheng et al (2008); Xiong et al (2015)) and voxel-based segmentations (e.g. Kirişli et al (2010)).…”

Benchmarking framework for myocardial tracking and deformation algorithms: An open access database

“…The approach described in Zheng et al (2008) was applied for joint detection and tracking of landmarks from LA views. It first identifies regions which are likely to contain the landmarks of interest and then applies machine learning based landmark detector to voxels within this region.…”

Automatic cardiac flow quantification on 3D volume color Doppler data

“…Automatic anatomy detection: In the first frame, the endocardial boundary of the left ventricle (LV), the mitral annulus, and the left ventricular outflow tract (LVOT) are detected using Marginal Space Learning (MSL) [8]. 2.…”

“…A 3D detector is learned to locate the pose, including the position X = (x, y, z), orientation θ = (α, β, γ) and scale S = (s x , s y , s z ), of the LV using the marginal space learning (MSL) approach [8]. The local deformations of the mitral annulus, LVOT, and myocardial boundaries are further estimated based on the posterior distribution p i (X|I) of each control point on the surface, which is learned using the steerable features and the probability boosting-tree (PBT) [9].…”

“…where F t is the steerable feature response [8], T t is the local image template, and λ is the weighting coefficient of the matching term. Given the resulting shapes X 1:t−1 from the previous t − 1 frames, the prediction term p( X t | Y 1:t−1 ) can be simplified as p( X t | X 1:t−1 ), which can be learned from the training data set as in [11].…”