Key result
DeepSurv model using aortic pressure during PCI predicts one-year STEMI mortality with ~0.88 C-index.
Why the study?
Accurate risk assessment in STEMI is essential for early identification of patients at increased risk of adverse outcomes following successful revascularization, building on prior work showing the predictive value of aortic pressure signals.
Does combining aortic pressure signals recorded during PPCI with survival models improve the prediction of 1-year mortality in STEMI patients?
Observational (n=600)
No
Does combining aortic pressure signals recorded during PPCI with survival models improve the prediction of 1-year mortality in STEMI patients?
Deep learning survival models utilizing aortic pressure signals recorded during primary PCI can accurately predict 1-year mortality in patients with STEMI.
Answers are grounded in this paper and cite their sources.
DeepSurv's high performance supports deep learning for STEMI mortality prediction; hypothesis-generating for aortic pressure integration, requiring prospective validation.
BACKGROUND: ST elevation myocardial infarction (STEMI) is a life-threatening condition, and is associated with significant mortality, especially in patients encountering cardiogenic shock. Accurate risk assessment in this population is essential for improving prognosis through the early identification of patients at increased risk of adverse outcomes following successful revascularization. Our previous work has demonstrated outcome predictive value of the aortic pressure (AP) signal. OBJECTIVES: The present study uses survival models and features primarily extracted from the AP signal, recorded during primary percutaneous coronary intervention (PPCI), to predict 1-year all-cause mortality. METHODS: A total of 600 STEMI patients treated with PPCI (64.2 ± 13.2 years; 171 [28.5%] females) were included in this single-center, retrospective study. Three survival models (Cox proportional hazards, DeepSurv, and random survival forest) were applied to predict all-cause mortality using AP-derived features. Then, using Shapley Additive Explanations (SHAP), we enhanced model interpretability and identified the most important features contributing to the best-performing model. RESULTS: DeepSurv outperformed the other two survival models, achieving an average concordance index (C-index) of 0.877, an integrated Brier score (IBS) of 0.047, and a mean time-dependent area under the curve (AUC) of 0.907. The SHAP method identified key features extracted from the AP signal, including systolic area, ejection systolic time, kurtosis, diastolic blood pressure, and the age-modified shock index. CONCLUSIONS: Combining AP signals recorded during PPCI with survival models significantly improved the prediction of 1-year mortality in the STEMI cohort.
No takes yet. Share an insight, caveat, or question.
Razavi et al. (2026) conducted an observational in ST elevation myocardial infarction (STEMI) (n=600). DeepSurv survival model using aortic pressure (AP) signal features vs. Cox proportional hazards and random survival forest models was evaluated on 1-year all-cause mortality (C-index 0.877). The DeepSurv survival model using aortic pressure signals during primary PCI predicted 1-year all-cause mortality in STEMI patients, achieving a C-index of 0.877 and an AUC of 0.907.
Synapse has enriched 10 related papers. Consider them for comparative context: