Related Experiment Video
Updated: Aug 27, 2025
10:25
Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
Published on: September 25, 2019
48.3K
Unsupervised domain adaptive myocardial infarction MRI classification diagnostics model based on target domain
Weifang Xie1, Yuhan Ding1, Zhifang Liao1
1School of Computer Science and Engineering, Central South University, Changsha 410000, China.
Computer Methods and Programs in Biomedicine
|October 2, 2022
Summary
This study introduces CardiacCN, an unsupervised domain adaptive MRI classification model to improve myocardial infarction diagnosis. The model enhances classification accuracy in unsupervised scenarios, aiding cardiologists in routine cardiac MRI examinations.
Area of Science:
- Medical Imaging and Artificial Intelligence
- Cardiovascular Disease Diagnosis
- Deep Learning for Medical Applications
Background:
- Myocardial infarction diagnosis using MRI is prone to errors due to circulatory system blockages.
- Current MRI classification methods for myocardial infarction lack efficiency and accuracy.
- Deep learning offers potential for developing computer-aided diagnostic tools for cardiac MRI.
Purpose of the Study:
- To develop an unsupervised domain adaptive MRI classification model for myocardial infarction.
- To improve the accuracy and reliability of myocardial infarction classification in cardiac MRI.
- To aid cardiologists in routine examinations by enhancing computer-aided diagnostic algorithms.
Main Methods:
- Proposed CardiacCN, an unsupervised MRI classification technique for myocardial infarction.
- Employed two distinct domain classifiers to achieve domain adaptation between different data fields.
- Utilized adversarial learning with resampling of target-domain confusion samples for improved classification.
Main Results:
- CardiacCN demonstrated improved performance on six domain adaptation tasks of the Sunnybrook Cardiac Dataset (SCD).
- Achieved an approximate 1.2% increase in mean target area myocardial infarction MRI classification accuracy.
- The model showed robustness to hyper-parameter variations and enhanced source domain predictor accuracy for target domain classification.
Conclusions:
- CardiacCN effectively addresses misclassification issues in target-domain myocardial infarction scans.
- The model fully utilizes implicit image classification information from the target domain.
- Improved knowledge transfer and classification accuracy for myocardial infarction in unsupervised clinical scans.
