Key result
Compact Bi-mLSTM model achieves ~0.92 macro-F1 score for real-time ECG delineation.
Why the study?
Accurate ECG waveform delineation underpins cardiac diagnosis and screening, yet the most accurate deep-learning models are too large for resource-constrained wearable and implantable systems.
Does a compact, uncertainty-aware Bi-mLSTM model with 2,522 parameters achieve accurate and calibrated real-time P, QRS, and T waveform delineation compared to expert annotations and larger models in subjects from the LUDB and patients with LVADs?
Does a compact, uncertainty-aware Bi-mLSTM model with 2,522 parameters achieve accurate and calibrated real-time P, QRS, and T waveform delineation compared to expert annotations and larger models in subjects from the LUDB and patients with LVADs?
A compact 2,522-parameter Bi-mLSTM model enables accurate, calibrated, and uncertainty-aware real-time ECG waveform delineation on resource-constrained edge devices, matching or exceeding models over 100 times larger.
Answers are grounded in this paper and cite their sources.
May support wearable ECG monitoring on edge devices; hypothesis-generating and requires prospective clinical validation before practice change.
Accurate electrocardiogram (ECG) waveform delineation underpins cardiac diagnosis and screening, yet the most accurate deep-learning models are too large for resource-constrained wearable and implantable systems. We present a compact, uncertainty-aware bidirectional matrix-memory LSTM (Bi-mLSTM) for per-sample P, QRS, and T delineation with only 2,522 parameters. The model operates on beat windows placed by automatic R-peak detection. It was selected and trained on the automatic annotations of the PhysioNet QT Database, frozen, and evaluated against expert annotations of the Lobachevsky University Database (LUDB, 200 subjects) and a clinical cohort of 25 patients with left ventricular assist devices (LVAD-DB). The tolerance-aware macro-F1 was 0.919 on LUDB and 0.875 on LVAD-DB (strict 0.850 and 0.761), the expected calibration error of the raw confidence was 0.054 and 0.041, and low confidence flagged the model’s own errors. Under tolerance-aware scoring, the model exceeded three published architectures with 0.3 to 1.4 million parameters, trained with the same windowing and procedure, on LUDB and matched the strongest of them on LVAD-DB with 119 times fewer parameters. Per record, model-derived QRS duration and QT exceeded the expert values by 12.7 and 11.3ms on LUDB, which supports trend monitoring rather than replacement of expert measurement. On an Arduino Portenta H7 microcontroller, a separately trained earlier version of the same architecture delineated a live recording in real time without quantization, with 128.6ms of processing per beat and 22 kB of memory. Accurate, calibrated, and uncertainty-aware ECG delineation is therefore feasible in real time on resource-constrained devices.
No takes yet. Share an insight, caveat, or question.
Lung et al. (2026) studied Electrocardiogram (ECG) waveform delineation in subjects from LUDB and patients with left ventricular assist devices (LVAD) (n=225). Compact bidirectional matrix-memory LSTM (Bi-mLSTM) model (2,522 parameters) vs. Expert ECG annotations and three published deep-learning architectures (0.3 to 1.4 million parameters) was evaluated on Tolerance-aware macro-F1 score for per-sample P, QRS, and T wave delineation. A compact 2,522-parameter Bi-mLSTM model achieved tolerance-aware macro-F1 scores of 0.919 on LUDB and 0.875 on LVAD-DB for real-time ECG delineation.
Synapse has enriched 10 related papers. Consider them for comparative context: