Phantom-Based Ultrasound-ECG Deep Learning Framework for Prospective Cardiac Computed Tomography.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiology

Background:

• Computed tomography angiography (CTA) gating relies on predicting cardiac quiescent periods (QPs) for diagnostic image quality.
• Current single-modality gating methods, often ECG-based, struggle with variable heart rates and arrhythmias.
• Optimizing CTA gating is crucial to avoid non-diagnostic scans due to motion artifacts.

Purpose of the Study:

• To develop and validate a novel multimodal deep learning framework for enhanced cardiac quiescent period (QP) prediction.
• To integrate ultrasound (US) and electrocardiography (ECG) data for more robust QP prediction in cardiac computed tomography angiography (CTA) gating.
• To improve the accuracy and reliability of CTA gating, particularly in patients with arrhythmias and variable heart rates.

Main Methods:

• A multimodal deep learning framework was developed, integrating a 3D convolutional neural network (CNN) for US data and an artificial neural network (ANN) for ECG data.
• The framework was validated using a dynamic heart motion phantom simulating diverse cardiac conditions, including arrhythmias.
• Performance was evaluated across various QP lengths, cardiac segments, and motion patterns to mimic real-world clinical scenarios.

Main Results:

• The multimodal US-ECG 3D CNN-ANN framework achieved 96.87% accuracy in QP prediction, significantly outperforming single-modality ECG-only gating (85.56%).
• The framework demonstrated superior performance in predicting QPs during arrhythmic conditions and for longer durations (100-200 ms).
• Accurate QP prediction was consistently achieved across different cardiac regions, with an average accuracy of 92% in clinically relevant echocardiographic views.

Conclusions:

• Combining US and ECG data via a multimodal deep learning framework significantly enhances QP prediction accuracy, especially under variable cardiac motion and in arrhythmic patients.
• This multimodal approach offers a promising solution for improving diagnostic scan rates in CTA by providing more reliable gating.
• The framework's robustness across various cardiac conditions suggests potential for widespread clinical adoption in improving cardiac imaging quality.