Introduction and Comparison of Novel Decentral Learning Schemes with Multiple Data Pools for Privacy-Preserving ECG Classification

Q: What is the clinical evidence from this study?

Population: ECG classification (n=43101). Intervention: Batch-wise sequential learning (decentralized machine learning) vs. Centralized machine learning. Primary outcome: Area under the receiver operator curve (AUROC) (-0.036 AUROC).

Martin Baumgartner; Sai Veeranki; Dieter Hayn; G. Schreier

doi:10.1007/s41666-023-00142-5

August 17, 2023Journal of Healthcare Informatics ResearchOpen Access

Introduction and Comparison of Novel Decentral Learning Schemes with Multiple Data Pools for Privacy-Preserving ECG Classification

View Full Paper

Ask AI

Bookmark

Share

Key result

Batch-wise sequential learning achieves near-centralized ECG classification with only a ~0.04 drop in AUROC.

-0.036 AUROC
n=43,101

Figures

Why the study?

Applying AI and machine learning in medicine is challenging due to privacy concerns and strict legal regulations, which could be addressed by methods that centralize knowledge instead of data.

Do decentralized machine learning algorithms achieve comparable ECG classification performance to centralized machine learning on distributed datasets?

Structured PICO

Do decentralized machine learning algorithms achieve comparable ECG classification performance to centralized machine learning on distributed datasets?

P

Population

43,101 ECG recordings from four distinct open-source datasets used to simulate a distributed learning setting for 12-lead ECG classification.

I

Intervention

Six decentralized machine learning algorithms (including standard federated learning, weighted federated learning, node-wise sequential learning, batch-wise sequential learning, regression ensemble, and weighted regression ensemble).

C

Comparator

Conventional, centralized machine learning (baseline model trained with all data pooled centrally).

O

Outcome

Classification performance measured by AUROC, accuracy, Jaccard score, F1 score, specificity, and sensitivity.surrogate

Main Result

Mean Difference: -0.036

Absolute Event Rate: 0.836% vs 0.872%

Decentralized machine learning schemes, particularly batch-wise sequential learning, can achieve ECG classification performance comparable to centralized models while preserving data privacy.

Limitations

Simulated scenario with only four data nodes rather than a true real-world distributed network.
Technical aspects of implementing the algorithms in a real-world application remain to be addressed.
Small datasets produced poor individual models, highlighting challenges with heterogeneous node sizes.
Technical aspects of implementing the algorithms in a real-world application need careful consideration.

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

May support privacy-preserving decentralized ECG classification; hypothesis-generating pending prospective clinical validation.

Abstract

Artificial intelligence and machine learning have led to prominent and spectacular innovations in various scenarios. Application in medicine, however, can be challenging due to privacy concerns and strict legal regulations. Methods that centralize knowledge instead of data could address this issue. In this work, 6 different decentralized machine learning algorithms are applied to 12-lead ECG classification and compared to conventional, centralized machine learning. The results show that state-of-the-art federated learning leads to reasonable losses of classification performance compared to a standard, central model (-0.054 AUROC) while providing a significantly higher level of privacy. A proposed weighted variant of federated learning (-0.049 AUROC) and an ensemble (-0.035 AUROC) outperformed the standard federated learning algorithm. Overall, considering multiple metrics, the novel batch-wise sequential learning scheme performed best (-0.036 AUROC to baseline). Although, the technical aspects of implementing them in a real-world application are to be carefully considered, the described algorithms constitute a way forward towards preserving-preserving AI in medicine.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Baumgartner et al. (2023) studied ECG classification (n=43,101). Batch-wise sequential learning (decentralized machine learning) vs. Centralized machine learning was evaluated on Area under the receiver operator curve (AUROC) (-0.036 AUROC). A novel batch-wise sequential learning scheme achieved the best decentralized performance for ECG classification, with only a minor loss in AUROC (-0.036) compared to a centralized baseline model.

synapsesocial.com/papers/6a204b9a89b72f34e0e71794 https://doi.org/10.1007/s41666-023-00142-5

View Full Paper

Ask AI

Bookmark

Share

Introduction and Comparison of Novel Decentral Learning Schemes with Multiple Data Pools for Privacy-Preserving ECG Classification

Figures

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Introduction and Comparison of Novel Decentral Learning Schemes with Multiple Data Pools for Privacy-Preserving ECG Classification

Figures

Structured PICO

Main Result

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading