SCAR-AWARE LATE MECHANICAL ACTIVATION DETECTION NETWORK FOR OPTIMAL CARDIAC RESYNCHRONIZATION THERAPY PLANNING

Area of Science:

• Cardiovascular Imaging
• Medical Image Analysis
• Computational Cardiology

Background:

• Accurate identification of late mechanical activation (LMA) regions is vital for effective cardiac resynchronization therapy (CRT) lead implantation.
• Current cardiac magnetic resonance (CMR) imaging methods for LMA detection may misidentify myocardial scar as delayed activation, potentially compromising CRT outcomes.
• Integrating scar information into LMA detection is essential for improving the precision of CRT planning.

Purpose of the Study:

• To develop and validate a scar-aware LMA detection network that simultaneously identifies myocardial scar and LMA regions.
• To prevent the misclassification of scar tissue as LMA, thereby enhancing the accuracy of LMA mapping.
• To improve the overall effectiveness of CRT planning through more precise LMA identification.

Main Methods:

• Integration of a pre-trained scar segmentation network (using late gadolinium enhancement - LGE CMRs) with a LMA detection network utilizing displacement encoding with stimulated echoes (DENSE) CMRs.
• Introduction of a novel scar-aware loss function to penalize false-positive LMA detections in scarred regions.
• Model training capability with or without paired LGE data, and inference without requiring LGE images, relying on strain data patterns.

Main Results:

• Demonstrated significantly improved LMA detection accuracy in subjects with and without myocardial scar.
• Successfully mitigated false-positive LMA detection in potential scar regions during inference.
• Validated the model's efficacy in scenarios with and without scar, showcasing robust performance.

Conclusions:

• The proposed scar-aware LMA detection network offers a significant advancement in identifying LMA regions for CRT.
• This approach enhances the accuracy of LMA mapping by effectively distinguishing scar tissue from delayed activation.
• The findings pave the way for improved CRT planning and potentially better patient outcomes.