A 10 Hz signal processing pipeline maintained or improved F1 scores relative to native-rate baselines in 10 of 12 scenarios, with gains up to +3.57% on severely noisy signals.
A 10 Hz sparse representation pipeline for QRS detection improves performance on noisy and complex ECG signals while significantly increasing computational throughput.
QRS Detection Algorithms Optimized for 10 Hz Input Rate Accurate, low-power QRS detection remains challenging at ultra-low sampling rates. We present a signal processing pipeline that operates at a 10 Hz input rate using morphology-aware extrema extraction, together with two structurally adapted detectors (MinEl, TiTen) derived from Elgendi and Pan‑Tompkins. Parameter optimization is performed via mixed‑integer surrogate optimization on a multi‑database training corpus. Prospective evaluation on six PhysioNet databases (2.8 million labeled beats) shows that the pipeline maintains or improves F1 relative to native‑rate baselines in 10 of 12 scenarios, with significant gains on severely noisy (NSTDB: +3.57 %) and morphologically complex signals (TWADB: +3.24 %). It reduces per‑patient F1 variance by up to 90 %, increases wall‑clock throughput by 1.6× to 871×, and achieves a maximum real‑time factor of 886,000 on a consumer laptop. A 10 Hz sparse representation that preserves local extrema is not merely viable but advantageous for robust, resource‑constrained beat detection. Source code available from: https://github.com/titusnkumara/TiTen-MinEl
Titus Jayarathna (Fri,) conducted a other in QRS detection. 10 Hz input rate signal processing pipeline (MinEl, TiTen detectors) vs. Native-rate baselines was evaluated on F1 score for QRS detection. A 10 Hz signal processing pipeline maintained or improved F1 scores relative to native-rate baselines in 10 of 12 scenarios, with gains up to +3.57% on severely noisy signals.