Ballistocardiography Signal Quality Assessment Using Machine Learning.

Area of Science:

• Biomedical Engineering
• Cardiovascular Physiology
• Machine Learning

Background:

• Ballistocardiography (BCG) is a non-invasive method for monitoring cardiac activity.
• BCG shows potential for magnetic resonance imaging (MRI) gating applications.
• Accurate heartbeat detection is vital for precise MR triggering.

Purpose of the Study:

• To develop an automated method for identifying Ballistocardiography signal quality.
• To enhance heartbeat detection accuracy for improved MR triggering.
• To classify BCG segments as high- or low-quality using machine learning.

Main Methods:

• Analyzed 21 statistical signal features of BCG data.
• Evaluated 14 machine learning models for BCG signal classification.
• Utilized ensemble classifiers, including the Extra Trees Classifier.

Main Results:

• The Extra Trees Classifier achieved 86.65% overall accuracy, 80.54% sensitivity, and 73.62% precision.
• Time-domain features demonstrated superior performance over frequency-domain features.
• Four ensemble classifiers exhibited the best performance in classifying BCG signal quality.

Conclusions:

• Automated BCG signal quality assessment improves heartbeat detection accuracy.
• This method enhances the robustness of BCG applications, particularly in cardiac MRI.
• The technique automates channel selection, reducing reliance on visual assessment and improving MRI data acquisition.