Key result
Custom Tp-e delineator yields accepted outputs in ~64% of ECGs but fails all performance criteria.
Why the study?
Does the custom Tp-e delineator accurately measure ECG intervals compared to a reader-derived human reference in ECGs from the PTB-XL+ dataset?
Observational (n=160)
Does the custom Tp-e delineator accurately measure ECG intervals compared to a reader-derived human reference in ECGs from the PTB-XL+ dataset?
A record-level audit in PTB-XL+ demonstrated that a custom Tp-e delineator failed all performance acceptance criteria and had substantial disagreement with human reference measurements, rendering it unsuitable for confirmatory clinical analyses.
Answers are grounded in this paper and cite their sources.
Automated Tp-e algorithms require rigorous validation before clinical use; this audit leaves open questions on reliable ECG interval measurement.
This archive provides the submission manuscript, algorithm specification, deidentified derived analysis tables, figures, analysis manifest, and core Python scripts for a record-level audit of automated ECG interval measurements in PTB-XL+. The custom Tp-e delineator returned accepted outputs for 102 of 160 ECGs and failed all four algorithm-performance acceptance criteria. Among accepted outputs, substantial disagreement with the reader-derived human reference remained. The archive separates explicit non-output from error among retained outputs and includes calibration-overlap sensitivity analyses. Interpretation boundary: the evaluated delineator must not be used for confirmatory clinical-effect analyses. The package does not establish clinical risk thresholds or treatment effects. No credentials, private sample-mapping key, local reader-source CSVs, or unnecessary raw-waveform copies are included. PTB-XL and PTB-XL+ remain subject to their upstream terms.
No takes yet. Share an insight, caveat, or question.
Ye et al. (2026) conducted an observational in Automated ECG interval measurements (n=160). Custom Tp-e delineator vs. Reader-derived human reference was evaluated on Accepted outputs and algorithm-performance acceptance criteria. The custom Tp-e delineator returned accepted outputs for 102 of 160 ECGs and failed all four algorithm-performance acceptance criteria with substantial disagreement against human reference.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: