Key result
Adaptive cardiotocography preprocessing improves artifact detection F-measure by ~21 points at 10% signal loss.
Why the study?
Missing values and artifacts in cardiotocography signals directly affect preprocessing outcomes, and fixed thresholds fail to account for local signal characteristics.
Does a signal quality-adaptive preprocessing algorithm improve artifact detection and reduce reconstruction error in cardiotocography signals compared to fixed thresholds?
Does a signal quality-adaptive preprocessing algorithm improve artifact detection and reduce reconstruction error in cardiotocography signals compared to fixed thresholds?
An adaptive preprocessing algorithm combining a signal quality vector with local median and MAD-based detection significantly improves artifact detection and reduces reconstruction error in cardiotocography signals.
Missing values and artifacts in cardiotocography signals directly affect the outcome of preprocessing, while fixed thresholds do not account for the local characteristics of the signal. This paper proposes an algorithm that combines a signal quality vector, adaptive artifact detection based on the local median and MAD, and the selection of a reconstruction strategy according to gap length and context quality. The architecture of its software module was designed, and a Python prototype was implemented. In a synthetic protocol based on reference segments from the CTU-UHB database, the F-measure of artifact detection at 10% signal loss increased from 0.75 to 0.96 on 520 test instances drawn from 157 test recordings. The reconstruction error on the same gap samples decreased from 1.94 to 1.77 bpm, and the error over all corrupted samples decreased from 4.55 to 2.19 bpm. The computation time per sample was several orders of magnitude shorter than the sampling interval.
No takes yet. Share an insight, caveat, or question.
Satimovich et al. (2026) studied Cardiotocography signal artifacts and missing values (n=157). Signal quality-adaptive preprocessing algorithm (combining signal quality vector, local median/MAD artifact detection, and context-based reconstruction) vs. Fixed-threshold preprocessing was evaluated on F-measure of artifact detection at 10% signal loss. An adaptive cardiotocography preprocessing algorithm increased artifact detection F-measure from 0.75 to 0.96 and lowered reconstruction error from 4.55 to 2.19 bpm at 10% signal loss.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: