Generative Data by β-Variational Autoencoders Help Build Stronger Classifiers: ECG Use Case

被引：1

作者：

Nademi, Yousef ^{[1
]}

Kalmady, Sunil V. ^{[1
,2
,3
]}

Sun, Weijie ^{[1
]}

Salimi, Amir ^{[1
]}

Hindle, Abram ^{[1
]}

Kaul, Padma ^{[2
,3
]}

Greiner, Russell ^{[1
,4
]}

机构：

[1] Univ Alberta, Dept Comp Sci, Edmonton, AB, Canada

[2] Univ Alberta, Dept Med, Canadian VIGOUR Ctr, Edmonton, AB, Canada

[3] Univ Alberta, Dept Med, Edmonton, AB, Canada

[4] Alberta Machine Intelligence Inst, Edmonton, AB, Canada

来源：

2023 19TH INTERNATIONAL SYMPOSIUM ON MEDICAL INFORMATION PROCESSING AND ANALYSIS, SIPAIM | 2023年

基金：

加拿大自然科学与工程研究理事会;

关键词：

Electrocardiogram; Machine Learning; Variational Autoencoder; Data Generation;

D O I：

10.1109/SIPAIM56729.2023.10373478

中图分类号：

R318 [生物医学工程];

学科分类号：

0831 ;

摘要：

We explore the challenge of learning models that use electrocardiogram (ECG) data to diagnose various cardiovascular diseases. Here, we explore whether classifiers trained on a dataset of real labeled ECGs, augmented with synthetic ECGs, can perform better than ones trained on unaugmented datasets. We first used a dataset of ECGs, each labelled with one or more of 15 diagnoses, from 244,077 patients to train an unsupervised beta-VAE model, that could generate time series of 12-lead ECG signals for each of the diagnoses. We then used this generative model to generate ECGs with the ST-segment Elevated (STE) abnormality, which we added to the public dataset of ECG abnormalities (n = 6877, over normal (Sinus Rhythm) and 8 different abnormalities) of China Physiological Signal Challenge 2018, and found a learner trained on this extended dataset performed better than one trained on only the original data on the targeted STE label but also enhanced its performance for the classification of 4 other labels.

引用

页数：7

共 21 条

[1] A survey on ECG analysis [J].