Deep learning detects heart failure with preserved ejection fraction using a baseline electrocardiogram
Problem
Heart failure with preserved ejection fraction (HFpEF) lacks typical ECG features and currently requires clinical, invasive, and laboratory exams to diagnose. The study tested whether AI could detect HFpEF from a baseline ECG.
Why it matters
HFpEF is described as a rapidly growing global health problem, so a non-invasive ECG-based detection tool could ease diagnosis.
What the AI did
A convolutional neural network was trained on ECG segments from 1884 patients to classify HFpEF vs controls per ESC criteria, then validated on an external prospective screening cohort of 203 volunteers.
Results
External validation AUC 0.80 (95% CI 0.74-0.86) for detecting HFpEF per ESC criteria; sensitivity figure truncated in provided abstract.
Impact
The externally validated CNN shows an ECG-based tool can aid HFpEF detection, though performance is moderate.
Record
| Venue | European Heart Journal - Digital Health |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Matthias Unterhuber, Karl‐Philipp Rommel, Karl‐Patrik Kresoja, Julia Anna Lurz, Jelena Kornej, Gerhard Hindricks, Markus R. Scholz, Hölger Thiele et al. |
| Citations | 31 (OpenAlex, at import time) |
| References | 8 |
| Model or method | Convolutional neural network (CNN) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1093/ehjdh/ztab081 |
| Open access | gold |
| Keywords | heart failure, HFpEF, ECG, deep learning, CNN, diagnostics |
Importance rating
significant — thousands to millions affected
Abstract calls HFpEF a rapidly growing global health problem; first CNN for non-invasive ECG-based HFpEF detection, externally validated.