Related Experiment Video
Updated: Dec 7, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.2K
Cardiac magnetic resonance image segmentation based on convolutional neural network
Duqiu Liu1, Zheng Jia2, Ming Jin3
1Department of Cardiology, the Fifth Affiliated Hospital of Southern Medical University, Guangzhou, China.
Computer Methods and Programs in Biomedicine
|September 25, 2020
Summary
This study introduces a novel cardiac MRI segmentation technique using convolutional neural networks and image saliency, achieving high accuracy for improved heart disease diagnosis. The method enhances visualization of heart structures for effective clinical assessment.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Cardiology
Background:
- Cardiac MRI segmentation is crucial for heart disease diagnosis.
- Traditional methods struggle with poorly defined boundaries in cardiac MRI.
- Irregular heart motion complicates accurate image segmentation.
Purpose of the Study:
- To develop an automated cardiac MRI segmentation technique.
- To improve the accuracy of heart structure segmentation.
- To enhance clinical diagnosis of heart conditions.
Main Methods:
- Utilized a convolutional neural network (CNN) for target area detection and region of interest (ROI) extraction.
- Employed image saliency to enhance heart tissue clarity within the ROI.
- Compared the CNN-based method with a traditional region growth algorithm.
- Trained and tested the model on cardiac MRI images from 85 patients.
Main Results:
- Achieved high segmentation accuracy: 93.14% for ventricles, 92.58% for septum, and 96.21% for the apex.
- Demonstrated superior performance compared to the region growth segmentation technique.
- The CNN and saliency approach effectively filtered out non-cardiac anatomical structures.
Conclusions:
- The proposed CNN and image saliency method enables automatic heart segmentation from cardiac MRI sequences.
- The technique effectively assists physicians in observing patient heart health.
- This automated segmentation method shows significant potential for clinical applications.
