Lost in Tracking: Uncertainty-guided Cardiac Cine MRI Segmentation at Right Ventricle Base

Yidong Zhao,Yi Zhang,Orlando Simonetti,Yuchi Han,Qian Tao

DOI: https://doi.org/10.1007/978-3-031-72114-4_40

2024-10-17

Abstract:Accurate biventricular segmentation of cardiac magnetic resonance (CMR) cine images is essential for the clinical evaluation of heart function. However, compared to left ventricle (LV), right ventricle (RV) segmentation is still more challenging and less reproducible. Degenerate performance frequently occurs at the RV base, where the in-plane anatomical structures are complex (with atria, valve, and aorta) and vary due to the strong interplanar motion. In this work, we propose to address the currently unsolved issues in CMR segmentation, specifically at the RV base, with two strategies: first, we complemented the public resource by reannotating the RV base in the ACDC dataset, with refined delineation of the right ventricle outflow tract (RVOT), under the guidance of an expert cardiologist. Second, we proposed a novel dual encoder U-Net architecture that leverages temporal incoherence to inform the segmentation when interplanar motions occur. The inter-planar motion is characterized by loss-of-tracking, via Bayesian uncertainty of a motion-tracking model. Our experiments showed that our method significantly improved RV base segmentation taking into account temporal incoherence. Furthermore, we investigated the reproducibility of deep learning-based segmentation and showed that the combination of consistent annotation and loss of tracking could enhance the reproducibility of RV segmentation, potentially facilitating a large number of clinical studies focusing on RV.

Image and Video Processing,Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

This paper attempts to solve the problems of inaccurate and non - reproducible segmentation of the basal part of the right ventricle (RV) in cardiac magnetic resonance (CMR) cine images. Specifically, compared with the left ventricle (LV), the segmentation of the right ventricle is more challenging and unstable, especially in the basal area, because the anatomical structure here is complex (including the atrium, valves and aorta), and varies greatly due to strong out - of - plane motion. These problems lead to the degradation of the segmentation performance at the RV base. To solve these problems, the authors propose the following strategies: 1. **Dataset improvement**: Supplement public resources by relabeling the RV base in the ACDC dataset, especially by refining the labels of the right ventricular outflow tract (RVOT), and these labels are completed under the guidance of expert cardiologists. 2. **Dual - encoder U - Net architecture**: Propose a new dual - encoder U - Net architecture, which uses temporal incoherence (i.e., "loss - of - tracking") to guide segmentation, especially in the case of out - of - plane motion. Characterize the loss - of - tracking through Bayesian uncertainty estimation, thereby improving the segmentation performance. The experimental results show that this method significantly improves the segmentation accuracy of the RV base and enhances the reproducibility of segmentation. In addition, the study also explores the reproducibility problem of deep - learning - based segmentation, and finds that the combination of consistent labeling and loss - of - tracking can further improve the reproducibility of RV segmentation, which is helpful for more reliable RV evaluation in future clinical studies.

Lost in Tracking: Uncertainty-guided Cardiac Cine MRI Segmentation at Right Ventricle Base

Segmentation of Cardiac Tissues and Organs for CCTA Images Based on a Deep Learning Model

Reliable Off-Resonance Correction in High-Field Cardiac MRI Using Autonomous Cardiac B0 Segmentation with Dual-Modality Deep Neural Networks.

DeU-Net 2.0: Enhanced deformable U-Net for 3D cardiac cine MRI segmentation

Simultaneous segmentation and motion recovery in 3d cardiac image analysis

Cardiac MRI Segmentation with Focal Loss Constrained Deep Residual Networks

Automatic right ventricular segmentation for cine cardiac magnetic resonance images based on a new deep atlas network

Fully Automatic initialization and segmentation of left and right ventricles for large-scale cardiac MRI using a deeply supervised network and 3D-ASM

FR-NET: Focal Loss Constrained Deep Residual Networks for Segmentation of Cardiac MRI.

Spatio-temporal Active Region Model for Simultaneous Segmentation and Motion Estimation of the Whole Heart

Spatiotemporal Strategies for Joint Segmentation and Motion Tracking from Cardiac Image Sequences

Right Ventricular Segmentation from Short- and Long-Axis MRIs via Information Transition

Feature Tracking Cardiac Magnetic Resonance via Deep Learning and Spline Optimization

Progress of Right Ventricle Segmentation from Short-Axis Images Acquired with Cardiac Cine MRI

Left Ventricle Segmentation Via Optical-Flow-Net from Short-Axis Cine MRI: Preserving the Temporal Coherence of Cardiac Motion

Automatic Segmentation of Right Ventricle in Cine Cardiac Magnetic Resonance Image Based on a Dense and Multi-Scale U-net Method

Fully Automatic Segmentation of Short-Axis Cardiac MRI Using Modified Deep Layer Aggregation.

Myocardial Segmentation of Cardiac MRI Sequences With Temporal Consistency for Coronary Artery Disease Diagnosis

Toward Accurate Cardiac MRI Segmentation With Variational Autoencoder-Based Unsupervised Domain Adaptation

A Generative Adversarial Model for Right Ventricle Segmentation

Towards Accurate Cardiac MRI Segmentation with Variational Autoencoder-Based Unsupervised Domain Adaptation