Cine cardiac MRI slice misalignment correction towards full 3D left ventricle segmentation

Abstract

Accurate segmentation of the left ventricle (LV) blood-pool and myocardium is required to compute cardiac function assessment parameters or generate personalized cardiac models for pre-operative planning of minimally invasive therapy. Cardiac Cine Magnetic Resonance Imaging (MRI) is the preferred modality for high resolution cardiac imaging thanks to its capability of imaging the heart throughout the cardiac cycle, while providing tissue contrast superior to other imaging modalities without ionizing radiation. However, there exists an inevitable misalignment between the slices in cine MRI due to the 2D + time acquisition, rendering 3D segmentation methods ineffective. A large part of published work on cardiac MR image segmentation focuses on 2D segmentation methods that yield good results in mid-slices, however with less accurate results for the apical and basal slices. Here, we propose an algorithm to correct for the slice misalignment using a Convolutional Neural Network (CNN)-based regression method, and then perform a 3D graph-cut based segmentation of the LV using atlas shape prior. Our algorithm is able to reduce the median slice misalignment error from 3.13 to 2.07 pixels, and obtain the blood-pool segmentation with an accuracy characterized by a 0.904 mean dice overlap and 0.56 mm mean surface distance with respect to the gold-standard blood-pool segmentation for 9 test cine MR datasets.

Improving cardiac MRI convolutional neural network segmentation on small training datasets and dataset shift: A continuous kernel cut approach

“…Zheng et al (2015) formulated CRF as a recurrent layer as a part of a single unified CNN. In medical image segmentation, researchers have investigated the combination of CNN with active contours (Rupprecht et al, 2016), level-sets (Avendi et al, 2016;Ngo et al, 2017), CRF (Dou et al, 2017;Wang et al, 2018), graph cut (Dangi et al, 2018;Li et al, 2019) and continuous max-flow (Guo et al, 2018).…”

“…In addition, we employed normalized cuts for feature clustering because of the unique properties of balanced feature partitioning without a shrinking bias, which fundamentally differ from Potts and dense CRF methods, with superior performance (Tang et al, 2018;Veksler, 2019). Dangi et al (2018) also employed deep learning, atlases, and discrete graph cut for cardiac MRI left ventricle segmentation. This approach (Dangi et al, 2018) employed a CNN for LV centre detection for slice misalignment correction, atlas generation and registration to a target image, and iterative discrete graph cut for refining the propagated atlas labels.…”

“…Dangi et al (2018) also employed deep learning, atlases, and discrete graph cut for cardiac MRI left ventricle segmentation. This approach (Dangi et al, 2018) employed a CNN for LV centre detection for slice misalignment correction, atlas generation and registration to a target image, and iterative discrete graph cut for refining the propagated atlas labels. Our approach substantially differs from the previous work in that: 1) we utilized CNN segmentation probability maps to initialize a continuous kernel cut model that is comprised of normalized cuts and continuous regularization; and 2) we devel-oped a way to efficiently optimize the high-order segmentation model through upper bound optimization and convex relaxation within a spatially continuous max-flow framework, and demonstrated guaranteed algorithm convergence.…”

CNN-Based Cardiac Motion Extraction to Generate Deformable Geometric Left Ventricle Myocardial Models from Cine MRI

“…We train a modified version of the U-Net model [14] to segment the cardiac chambers, namely LV blood-pool, LV myocardium and RV blood-pool, from 2D cardiac MR images. We identify the LV bloodpool center, i.e., the centroid of the predicted segmentation mask and stack the 2D cardiac MR slices collinearly to obtain slice misalignment corrected 3D images [7,19].…”

A Convolutional Neural Network-based Deformable Image Registration Method for Cardiac Motion Estimation from Cine Cardiac MRI Images

“…To reduce artifacts caused by inherent slice misalignments during cine CMR image acquisition, we leverage the slice misalignment correction method presented by Dangi et al [6]. We train a modified version of the U-Net model [7] inspired from Isensee et al [8], to segment the cardiac chambers (LV blood-pool, LV myocardium and RV blood-pool) from 2D CMR slices.…”