PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
July 19, 201887 citations

Detection of Paroxysmal Atrial Fibrillation using Attention-based Bidirectional Recurrent Neural Networks

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Does an attention-based deep learning framework improve the detection of paroxysmal atrial fibrillation from ECG recordings compared to baseline models?

Population

2850 patients with 24-hour Holter Electrocardiogram recordings from the University of Virginia heart station.

Comparison

Attention-based deep learning framework applied… vs Baseline models

Design

Other

Key result

An attention-based bidirectional recurrent neural network detected paroxysmal atrial fibrillation from 24-hour Holter ECG recordings with an AUC of 0.94, exceeding baseline models.

Authors

SSSupreeth P. ShashikumarASAmit ShahGCGari D. Clifford

Discussion

Loading...

Member takes

Overview

May support AI integration in ambulatory AF monitoring; leaves open prospective outcome validation before practice change.

Structured PICO

Does an attention-based deep learning framework improve the detection of paroxysmal atrial fibrillation from ECG recordings compared to baseline models?

P
Population
2850 patients with 24-hour Holter Electrocardiogram (ECG) recordings from the University of Virginia heart station.
I
Intervention
Attention-based deep learning framework (deep convolutional neural network for image-based feature extraction followed by a bidirectional recurrent neural network with an attention layer) applied to time-frequency representations of 30-second recording windows over 10-minute data segments.
C
Comparator
Baseline models
O
Outcome
Detection of paroxysmal atrial fibrillation episodes (measured by Area Under the Curve [AUC])surrogate

An attention-based deep learning framework achieved high accuracy (AUC 0.94) in detecting paroxysmal atrial fibrillation from Holter ECGs, demonstrating potential for integration into wearable sensors for long-term monitoring.

Main Result

Effect estimate: AUC 0.94

Cite This Study

Shashikumar et al. (2018) studied Paroxysmal Atrial Fibrillation (n=2,850). Attention-based bidirectional recurrent neural network vs. Baseline models was evaluated on Detection of paroxysmal AF episodes (AUC 0.94). An attention-based bidirectional recurrent neural network detected paroxysmal atrial fibrillation from 24-hour Holter ECG recordings with an AUC of 0.94, exceeding baseline models.

synapsesocial.com/papers/6a11d62d71528255b2219f8chttps://doi.org/10.1145/3219819.3219912
View Full Paper
Ask AI
Bookmark
Share