Dynamic distribution entropy analysis via ultrafast intracardiac echocardiography for monitoring of cardiac radiofrequency ablation.

Abstract

To enable reliable, noise-robust monitoring of lesion formation during radiofrequency (RF) ablation with intracardiac echocardiography (ICE), this study introduces Dynamic Distribution Entropy (DDE), which quantifies temporal-spatial changes in backscatter statistics while suppressing ablation-induced interference. DDE integrates ultrafast plane-wave acquisition with singular value decomposition clutter filtering to stabilize entropy estimates. We evaluated DDE in (i) simulations modeling ablation-driven scatterer-size changes and (ii) ex-vivo porcine hearts imaged by ICE during RF ablation. DDE was compared with typical Shannon entropy, k-nearest neighbor entropy, cumulative residual entropy, and horizontally normalized Shannon entropy. Metrics included structural similarity (SSIM), intersection-over-union (IoU), and lesion-size agreement versus optical ground truth. In simulations, DDE and CRE closely tracked dynamic scatterer evolution, yielding highest SSIM over frames (0.96). In ex-vivo experiments, DDE demonstrated the most accurate lesion-size estimation under RF-ON conditions, exhibiting the lowest bias (4.85 mm²), standard deviation (1.53 mm²), and root mean square error (5.08 mm²) among all evaluated entropy-based methods. Lesion sizes derived from DDE exhibited the best agreement with optical measurements. DDE provides robust and accurate intraoperative monitoring of lesion formation in ICE guided RF ablation, outperforming conventional entropy imaging under ablation-related noise. DDE offers a practical, noise-resistant quantitative ultrasound biomarker for real-time lesion assessment, supporting decision-making during ICE-guided cardiac ablation without requiring changes to clinical workflow.