Scientific Sessions 2025  /  Integrating AI with ECG and Physiologic Signals for Multimodal Precision Health  /  Enhanced Two Dimensional Warping For Fast, Accurate, and Robust Segmentation of Continuous Ambulatory ECG Recordings
Logo

American Heart Association

  55
  0

Final ID: MP1515

Enhanced Two Dimensional Warping For Fast, Accurate, and Robust Segmentation of Continuous Ambulatory ECG Recordings

Abstract Body (Do not enter title and authors here): Introduction: Accurate segmentation of long-term electrocardiogram (ECG) signals is essential for diagnosing arrhythmias and monitoring cardiac health. While traditional methods, such as wavelet-based approaches, offer reasonable performance in controlled environments, they often fail to maintain reliability in real-world applications due to noise, beat variability, and morphological changes. Two-dimensional warping (2DW), which aligns annotated templates to individual beats through geometry-based transformation, has shown promise for high-precision ECG segmentation. However, its application to long-term recordings remains limited by its vulnerability to data drift and high computational costs.
Methods: We developed an enhanced 2DW framework that addresses both data drift and computational inefficiency through two core innovations: Dynamic Template Adaptation, which combines automatic beat detection, norm-one-based beat normalization, beat-to-template similarity tracking, and adaptive windowing change detection (ADWIN) to continuously update the template in response to data drift; and Multi-Point Iterative Warping, which iteratively transforms the template signal simultaneously at two, three or four reference points improving alignment accuracy while significantly reducing computational load. The enhanced 2DW framework was evaluated on 104 manually annotated, 15-minute, two-channel ECG recordings. Performance was benchmarked against the original 2DW. Metrics included sensitivity, measurement error, and processing time.
Results: The enhanced 2DW framework achieved slightly higher sensitivity compared to the original 2DW (95.6% vs 94.8%). It also demonstrated improved robustness, with reduced variability in Q-onset and T-offset detection. QT interval measurement accuracy was significantly enhanced over the original 2DW (P=0.039). In terms of computational efficiency, the enhanced 2DW achieved a fivefold speedup, reducing the average processing time from 171.44 seconds to 31.35 seconds per sample on an 8-core Apple M2 processor.
Conclusions: By combining dynamic template adaptation and optimized warping, our enhanced 2DW framework delivers improved accuracy, speed, and resilience to noise and arrhythmias in long-term ECG recordings. These advancements address critical challenges in real-time applications such as ambulatory cardiac monitoring, offering a scalable solution for remote patient care in resource-constrained environments.
  • Cachi, Paolo G.  ( Military Cardiovascular Outcomes Research , Bethesda , Maryland , United States )
  • Heroy, Andrew  ( Military Cardiovascular Outcomes Research , Bethesda , Maryland , United States )
  • Arnold, Thomas  ( Military Cardiovascular Outcomes Research , Bethesda , Maryland , United States )
  • Ehat, Nathaniel  ( Military Cardiovascular Outcomes Research , Bethesda , Maryland , United States )
  • Haigney, Mark  ( Military Cardiovascular Outcomes Research , Bethesda , Maryland , United States )
  • Solhjoo, Soroosh  ( F. Edward Hébert School of Medicine , Bethesda , Maryland , United States )
    • Author Disclosures:
    Paolo G. Cachi: DO NOT have relevant financial relationships | Andrew Heroy: DO NOT have relevant financial relationships | Thomas Arnold: No Answer | Nathaniel Ehat: No Answer | Mark Haigney: DO NOT have relevant financial relationships | Soroosh Solhjoo: No Answer
Meeting Info:

Scientific Sessions 2025

2025

New Orleans, Louisiana

Session Info:

Integrating AI with ECG and Physiologic Signals for Multimodal Precision Health

Sunday, 11/09/2025 , 09:15AM - 10:30AM

Moderated Digital Poster Session

More abstracts on this topic:
A large-scale multi-view deep learning-based assessment of left ventricular ejection fraction in echocardiography

Jing Linyuan, Metser Gil, Mawson Thomas, Tat Emily, Jiang Nona, Duffy Eamon, Hahn Rebecca, Homma Shunichi, Haggerty Christopher, Poterucha Timothy, Elias Pierre, Long Aaron, Vanmaanen David, Rocha Daniel, Hartzel Dustin, Kelsey Christopher, Ruhl Jeffrey, Beecy Ashley, Elnabawi Youssef

Advanced Diagnosis of Hypertrophic Cardiomyopathy with AI-ECG and Differences Based on Race and Subtype

Lewontin Myra, Perry Allison, Amos Kaitlyn, Ayers Michael, Kaplan Emily, Bilchick Kenneth, Barber Anita, Bivona Derek, Kramer Christopher, Parrish Anna, Mcclean Karen, Thomas Matthew

4364370_File000000.jpg
4364370_File000000.jpg
You have to be authorized to contact abstract author. Please, Login
Not Available