Key result
Deep-learning feature relevance and univariate statistical area under the curve showed poor agreement (ρ=0.18, P=0.25) in identifying ECG features contributing to heart failure detection.
Why the study?
Assessing feature contributions to diagnosis is commonly done by statistical analysis, but comparisons with deep learning for ECG-based heart failure detection have not been evaluated.
Does deep learning feature contribution assessment correlate with statistical analysis for ECG-based heart failure detection in patients with prior myocardial infarction?
Observational (n=58)
No
Does deep learning feature contribution assessment correlate with statistical analysis for ECG-based heart failure detection in patients with prior myocardial infarction?
Deep learning and statistical analysis identify different ECG features as important for detecting heart failure, highlighting discordance between the two methodological approaches.
Discordance between deep-learning and statistical ECG feature ranking for heart failure detection warrants cautious interpretation of either method; challenges assumptions of methodological agreement.
Assessing feature contributions to a specific diagnosis is commonly done by statistical analysis. In the context of heart failure (HF) diagnosis from the electrocardiogram (ECG), this work compares feature contributions assessed by deep learning with those obtained by statistical analysis. Data consists of ECG pairs (baseline and follow-up) from patients with a history of myocardial infarction. When the follow-up ECG was made, controls patients had remained stable, while cases patients had developed HF. The 42 features that characterized each ECG served as inputs of a deep-learning neural network (NN) created by our Repeated Structuring & Learning Procedure. Subject-specific feature ranking was obtained from the local-interpretable model-agnostic explanatory algorithm and processed to obtain feature relevances (FR). Additionally, 42 areas under the curve (AUC) by univariate statistical analysis were obtained. FR and AUC were compared by Pearson's correlation coefficient (p). After training, the NN had a 99% classification performance. FR ranged from 0.32 to 4.47; AUC ranged from 23% to 82%. Correlation analysis yielded no significant association between AUC and FR (ρ=0.18, P-value =0.25). Deep-learning and statistical-analysis feature contributions to HF detection were discordant. Further studies will investigate which of the two approaches better reflects clinical interpretation.
No takes yet. Share an insight, caveat, or question.
Sbrollini et al. (2022) conducted an observational in Heart Failure (n=58). Deep learning feature relevance vs. Univariate statistical analysis (AUC) was evaluated on Agreement between deep-learning feature relevance (FR) and statistical analysis (AUC) evaluated by Pearson's correlation coefficient (ρ=0.18, p=0.25). Deep-learning feature relevance and univariate statistical area under the curve showed poor agreement (ρ=0.18, P=0.25) in identifying ECG features contributing to heart failure detection.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: