August 4, 2023IEEE Sensors Journal

The proposed model achieved state-of-the-art performance across all four datasets, with up to an 8% F1 score gain compared to benchmark models.

View Full Paper

Ask AI

Bookmark

Why the study?

Existing deep learning models for ECG anomaly detection learn from relatively long signals and are heavily parameterized, requiring large time and computational resources during training.

Population

Single- and 12-lead ECG signals from four datasets

Comparison

DConv-LSTM-Net vs two existing benchmark models

Design

Model development and subject-independent ten-fold cross-validation study

Authors

TDTheekshana Dissanayake TFTharindu Fernando SDSimon Denman

Discussion

Member takes

Overview

May enhance research ECG analysis tools; leaves open clinical validation before practice adoption.

Structured PICO

Population

ECG datasets including the PhysioNet atrial fibrillation (AF) challenge dataset, the China Physiological challenge, the PTB-XL repository from PhysioNet, and the Georgia dataset

Intervention

DConv-LSTM-Net (a novel deep learning architecture exploiting dilated convolution layers and a recurrent component)

Comparator

Two existing benchmark models

Outcome

F1 score for ECG anomaly detectionsurrogate

A novel deep learning architecture (DConv-LSTM-Net) improves ECG anomaly detection performance and offers an explainable solution capable of learning from short single- and 12-lead ECG segments.

Cite This Study

Dissanayake et al. (2023) studied this question.

synapsesocial.com/papers/6a72a58ca2d7cf2e39c2f7e9 https://doi.org/10.1109/jsen.2023.3300752

View Full Paper

Ask AI

Bookmark

August 4, 2023IEEE Sensors Journal

The proposed model achieved state-of-the-art performance across all four datasets, with up to an 8% F1 score gain compared to benchmark models.

View Full Paper

Ask AI

Bookmark

Why the study?

Existing deep learning models for ECG anomaly detection learn from relatively long signals and are heavily parameterized, requiring large time and computational resources during training.

Population

Single- and 12-lead ECG signals from four datasets

Comparison

DConv-LSTM-Net vs two existing benchmark models

Design

Model development and subject-independent ten-fold cross-validation study

Authors

TDTheekshana Dissanayake TFTharindu Fernando SDSimon Denman

Discussion

Member takes

Overview

May enhance research ECG analysis tools; leaves open clinical validation before practice adoption.

Structured PICO

Population

ECG datasets including the PhysioNet atrial fibrillation (AF) challenge dataset, the China Physiological challenge, the PTB-XL repository from PhysioNet, and the Georgia dataset

Intervention

DConv-LSTM-Net (a novel deep learning architecture exploiting dilated convolution layers and a recurrent component)

Comparator

Two existing benchmark models

Outcome

F1 score for ECG anomaly detectionsurrogate

A novel deep learning architecture (DConv-LSTM-Net) improves ECG anomaly detection performance and offers an explainable solution capable of learning from short single- and 12-lead ECG segments.

Cite This Study

Dissanayake et al. (2023) studied this question.

synapsesocial.com/papers/6a72a58ca2d7cf2e39c2f7e9 https://doi.org/10.1109/jsen.2023.3300752

View Full Paper

Ask AI

Bookmark