Heart failure & cardiomyopathies
Electrocardiograph Analysis for Risk Assessment of Heart Failure with Preserved Ejection Fraction: A Deep Learning Model
Authors' conclusion
The DLM offers an accurate, cost-effective tool for HFpEF risk assessment
Study context
- Population
- Patients undergoing ECG and invasive LV catheterisation; Heart failure with preserved ejection fraction (HFpEF) high risk left ventricular end-diastolic pressure (LVEDP) > 12 mmHg vs low risk LVEDP ≤ 12 mmHg
- Participants
- 238 vs 117
- Comparison
- ECG-based deep learning model for HFpEF risk vs invasive LVEDP reference
- Cohort An = 238
- Cohort Bn = 117
- Outcome
- Diagnostic performance of the DLM for HFpEF risk (LVEDP and clinical symptoms)
Methodological detailsprospective validation cohort
- Prospective validation cohort
Performance
As reported in the abstract
DLM performance for HFpEF risk
| Test or model | Sensitivity | Specificity | Accuracy |
|---|---|---|---|
| Cohort A | — not stated in the abstract | — not stated in the abstract | 78% |
| Cohort B | 71.7% | 71.9% | 78% |
Sensitivity
Cohort A— not stated in the abstract
Cohort B71.7%
Specificity
Cohort A— not stated in the abstract
Cohort B71.9%
Accuracy
Cohort A78%
Cohort B78%
— not stated in the abstract
Further findings
Characteristics of DLM-identified high-risk vs low-risk HFpEF (Cohort B)
Diabetes
BMI DLM high risk 25.92 kg/cm² · DLM low risk 24.22 kg/cm²
CCB use
BNP DLM high risk 22 pg/mL · DLM low risk 20 pg/mL
E/E' DLM high risk 8.25 · DLM low risk 8.5
- Characteristics of Deep Learning Model (DLM)-identified high-risk vs low-risk HFpEF (Cohort B): Diabetes P < 0.01
- Characteristics of DLM-identified high-risk vs low-risk HFpEF (Cohort B): BMI P < 0.01
- Characteristics of DLM-identified high-risk vs low-risk HFpEF (Cohort B): Calcium channel blockers (CCB) P < 0.01
- Characteristics of DLM-identified high-risk vs low-risk HFpEF (Cohort B): BNP P = 0.71
- Characteristics of DLM-identified high-risk vs low-risk HFpEF (Cohort B): E/E' P = 0.66
Original publication
Electrocardiograph Analysis for Risk Assessment of Heart Failure with Preserved Ejection Fraction: A Deep Learning Model. ESC Heart Fail. Published 27 October 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →