Back

Detection of multiple cardiac abnormalities using Convolution, Positional Encoder and Transformer on 12-lead ECG recordings

Ford, A.; Lan, J.; Ng, K.

2024-06-17 health informatics
10.1101/2024.06.17.24309016
Show abstract

ObjectivesFirstly, we aimed to develop a system capable of detecting multiple cardiac abnormalities simultaneously from 12-lead ECG recordings. Secondly, we tried to improve the detection by analyzing the relationship between imbalanced datasets and optimal classification thresholds. MethodsA novel fusion of Convolution Positional Encoder and Transformer Encoder was used to solve the multi-label classification problem. We used a proper evaluation metric called area under the precision-recall curve (AUPRC) that enabled us to analyze the precision-recall trade-off and find the optimal thresholds. ResultsHaving outperformed other popular deep networks, the model achieved the highest AUPRC of 0.96 and f1-score of 0.90 on 42511-sample datasets. We also found the negative correlation coefficient of -0.68 be-tween optimal thresholds and the proportion of positive samples. Significance &ConclusionThis study compared the performances of different deep learning architectures on a medical problem and showed the potential of advanced techniques in capturing spatial, temporal features alongside attention mechanisms. It also introduced how to reduce the impact of imbalanced datasets and find optimal classification thresholds.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

1
Computers in Biology and Medicine
based on 39 papers
Top 0.1%
11.1%
2
BMC Medical Informatics and Decision Making
based on 36 papers
Top 0.7%
11.1%
3
Frontiers in Artificial Intelligence
based on 11 papers
Top 0.1%
11.0%
4
PLOS Digital Health
based on 88 papers
Top 2%
7.5%
5
Scientific Reports
based on 701 papers
Top 36%
5.0%
6
JMIR Medical Informatics
based on 16 papers
Top 1%
4.4%
50% of probability mass above
7
IEEE Journal of Biomedical and Health Informatics
based on 14 papers
Top 0.2%
4.4%
8
International Journal of Medical Informatics
based on 25 papers
Top 2%
2.9%
9
Journal of Biomedical Informatics
based on 37 papers
Top 2%
2.8%
10
Informatics in Medicine Unlocked
based on 11 papers
Top 0.5%
2.8%
11
IEEE Access
based on 11 papers
Top 0.5%
2.5%
12
European Heart Journal - Digital Health
based on 15 papers
Top 1%
2.4%
13
PLOS ONE
based on 1737 papers
Top 86%
2.3%
14
Computer Methods and Programs in Biomedicine
based on 12 papers
Top 0.4%
2.3%
15
Journal of Medical Internet Research
based on 81 papers
Top 9%
1.6%
16
Patterns
based on 15 papers
Top 1%
1.6%
17
Sensors
based on 18 papers
Top 2%
1.6%
18
npj Digital Medicine
based on 85 papers
Top 11%
1.3%
19
JAMIA Open
based on 35 papers
Top 5%
1.3%
20
Frontiers in Physiology
based on 18 papers
Top 2%
1.3%
21
Journal of the American Medical Informatics Association
based on 53 papers
Top 5%
1.3%
22
Frontiers in Cardiovascular Medicine
based on 33 papers
Top 5%
1.3%
23
JMIR Formative Research
based on 31 papers
Top 4%
1.3%
24
Frontiers in Digital Health
based on 18 papers
Top 4%
0.8%
25
Biology Methods and Protocols
based on 19 papers
Top 3%
0.8%
26
JMIR Public Health and Surveillance
based on 45 papers
Top 14%
0.7%
27
BMC Medical Research Methodology
based on 41 papers
Top 6%
0.7%