Regional heart motion abnormality detection via multiview fusion

Area of Science:

• Medical Imaging
• Cardiology
• Image Processing

Background:

• Accurate detection of regional heart motion abnormalities is crucial for diagnosing cardiac conditions.
• Existing methods often rely solely on short-axis cardiac MR images, potentially missing vital information.
• Integrating data from multiple imaging planes can offer a more comprehensive assessment of myocardial function.

Purpose of the Study:

• To develop and evaluate a novel multiview fusion technique for improved detection of regional heart motion abnormalities in cine cardiac MR images.
• To leverage information from both short-axis and long-axis views (2-, 3-, and 4-chamber) for a more robust analysis.
• To enhance the accuracy of classifying 17 standardized left ventricular segments according to American Heart Association guidelines.

Main Methods:

• Nonrigid image registration to obtain initial myocardial point correspondences within each sequence.
• 3D mapping and tracking of myocardial points using the Unscented Kalman Filter (UKS).
• A maximum likelihood-based track-to-track fusion approach to integrate data from multiple views.
• Classification of abnormal myocardial regions using Shannon's differential entropy and a naive Bayes classifier.

Main Results:

• The proposed multiview fusion method achieved an area-under-the-curve (AUC) of 95.9% in detecting abnormal myocardial function.
• This represents a significant improvement of 3.8% compared to methods using only short-axis images.
• Evaluations on 480 regional myocardial segments from 30 subjects demonstrated the efficacy of the fusion approach against radiologist ground truth.

Conclusions:

• Multiview fusion of cine cardiac MR images significantly enhances the accuracy of regional heart motion abnormality detection.
• The integration of long-axis views alongside short-axis views provides a more comprehensive assessment of left ventricular function.
• This advanced technique offers a promising tool for improved diagnosis and management of cardiac diseases.