Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
October 26, 2025SensorsOpen Access

Hybrid FusionViT model achieves ~86% accuracy for ECG arrhythmia detection.

View Full Paper
Ask AI
Bookmark
Share

Population

ECG signals from the PhysioNet Challenge 2017 dataset

Comparison

Scalograms fused with scattering and statistical features vs scalograms alone across diverse deep learning architectures

Design

Machine learning model development and validation study

Key result

A hybrid FusionViT architecture combining scalograms with scattering and statistical features achieved an accuracy of 0.8623 and F1-score of 0.8528 for ECG arrhythmia detection.

Authors

UTUkesh ThapaBPBipun Man PatiATAttaphongse Taparugssanagorn

Discussion

Loading...

Member takes

Overview

Enables potential wearable ECG arrhythmia detection; leaves open prospective clinical validation before adoption.

Structured PICO

P
Population
ECG signals from the PhysioNet Challenge 2017 dataset
I
Intervention
Progressive deep learning framework combining time-frequency representations (scalograms) with complementary hand-crafted features (scattering and statistical features) using architectures like FusionViT and Fusion ResNet-18
C
Comparator
Pure image-based ECG analysis (scalograms alone) and various deep learning architectures (SimpleCNN, ResNet-18, CNNTransformer, ViT)
O
Outcome
Accuracy and F1-score for ECG rhythm classificationsurrogate

Main Result

Absolute Event Rate: 0.8623% vs 0.859%

A hybrid wavelet-deep learning architecture combining visual and statistical signal features achieves high accuracy and efficiency for ECG arrhythmia detection, suitable for wearable devices.

Limitations

  • Scalograms alone showed variability across folds
  • Trade-off between efficiency and fine-grained temporal resolution

Cite This Study

Thapa et al. (2025) studied Arrhythmia. Hybrid wavelet-deep learning architecture (FusionViT) vs. Standard deep learning architectures (ViT, ResNet-18, SimpleCNN, CNNTransformer) was evaluated on Accuracy for ECG rhythm classification. A hybrid FusionViT architecture combining scalograms with scattering and statistical features achieved an accuracy of 0.8623 and F1-score of 0.8528 for ECG arrhythmia detection.

synapsesocial.com/papers/6a1786137afe20c06351e05dhttps://doi.org/10.3390/s25216590
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Hybrid Deep Learning Model for Scalogram-Based ECG Classification of Cardiovascular Diseases2025 · 2 citations
  2. 2Hybrid Deep Learning and Discrete Wavelet Transform-Based ECG Biometric Recognition for Arrhythmic Patients and Healthy Controls2023 · 12 citations
  3. 3A deep learning framework for arrhythmia classification: mitigating data leakage and class imbalance in ECG analysis2026
  4. 4Arrhythmias Detection using ECG with Deep Learning and Superlet Transform2025
  5. 5Transfer Learning-Based Electrocardiogram Classification Using Wavelet Scattered Features2023 · 17 citations