Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Fully Convolutional Neural Network for Cardiac Segmentation in Short-Axis MRI

Phi Vu Tran

TL;DR

The paper tackles automated segmentation of both left and right ventricles in short-axis cine MRI using a fully convolutional neural network trained end-to-end for per-pixel labeling. It demonstrates state-of-the-art or competitive performance across multiple public datasets (Sunnybrook, LVSC, RVSC) with fast GPU-enabled inference and effective transfer learning to adapt to limited data. The work highlights the scalability and practicality of FCN-based cardiac segmentation while acknowledging persistent challenges at apical/basal slices and the potential gains from larger annotated datasets. Overall, it establishes FCN-based semantic segmentation as a powerful tool for automated, large-scale cardiac ventricle delineation in clinical workflows.

Abstract

Automated cardiac segmentation from magnetic resonance imaging datasets is an essential step in the timely diagnosis and management of cardiac pathologies. We propose to tackle the problem of automated left and right ventricle segmentation through the application of a deep fully convolutional neural network architecture. Our model is efficiently trained end-to-end in a single learning stage from whole-image inputs and ground truths to make inference at every pixel. To our knowledge, this is the first application of a fully convolutional neural network architecture for pixel-wise labeling in cardiac magnetic resonance imaging. Numerical experiments demonstrate that our model is robust to outperform previous fully automated methods across multiple evaluation measures on a range of cardiac datasets. Moreover, our model is fast and can leverage commodity compute resources such as the graphics processing unit to enable state-of-the-art cardiac segmentation at massive scales. The models and code are available at https://github.com/vuptran/cardiac-segmentation

A Fully Convolutional Neural Network for Cardiac Segmentation in Short-Axis MRI

TL;DR

The paper tackles automated segmentation of both left and right ventricles in short-axis cine MRI using a fully convolutional neural network trained end-to-end for per-pixel labeling. It demonstrates state-of-the-art or competitive performance across multiple public datasets (Sunnybrook, LVSC, RVSC) with fast GPU-enabled inference and effective transfer learning to adapt to limited data. The work highlights the scalability and practicality of FCN-based cardiac segmentation while acknowledging persistent challenges at apical/basal slices and the potential gains from larger annotated datasets. Overall, it establishes FCN-based semantic segmentation as a powerful tool for automated, large-scale cardiac ventricle delineation in clinical workflows.

Abstract

Automated cardiac segmentation from magnetic resonance imaging datasets is an essential step in the timely diagnosis and management of cardiac pathologies. We propose to tackle the problem of automated left and right ventricle segmentation through the application of a deep fully convolutional neural network architecture. Our model is efficiently trained end-to-end in a single learning stage from whole-image inputs and ground truths to make inference at every pixel. To our knowledge, this is the first application of a fully convolutional neural network architecture for pixel-wise labeling in cardiac magnetic resonance imaging. Numerical experiments demonstrate that our model is robust to outperform previous fully automated methods across multiple evaluation measures on a range of cardiac datasets. Moreover, our model is fast and can leverage commodity compute resources such as the graphics processing unit to enable state-of-the-art cardiac segmentation at massive scales. The models and code are available at https://github.com/vuptran/cardiac-segmentation

Paper Structure

This paper contains 15 sections, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Previous Work
  3. Left Ventricle Segmentation
  4. Right Ventricle Segmentation
  5. Fully Convolutional Neural Networks
  6. Experimental Framework
  7. Datasets
  8. Data Preparation and Augmentation
  9. FCN Architecture
  10. Training Protocol
  11. Transfer Learning
  12. Empirical Evaluation
  13. Metrics
  14. Results and Analysis
  15. Conclusion

Figures (4)

  • Figure 1: A schematic of our proposed fully convolutional neural network architecture. Acronyms: ReLU -- Rectified Linear Unit; MVN -- Mean-Variance Normalization.
  • Figure 2: FCN segmentation result of an example test case in the Sunnybrook dataset for both ED and ES phases. Colors: red -- endocardium; green -- epicardium.
  • Figure 3: FCN segmentation result of an example test case in the RVSC dataset for both ED and ES phases. Colors: red -- endocardium; green -- epicardium.
  • Figure 4: Examples of poor FCN segmentation on difficult apical slices having ambiguous or imperceptible object boundaries. Cropped and zoomed in for better viewing. Colors: red -- endocardium; green -- epicardium.