Related Experiment Video
Updated: Aug 25, 2026
Human Fetal Blood Flow Quantification with Magnetic Resonance Imaging and Motion Compensation
Published on: January 7, 2021
Fourier Evaluation of Tracings and Acidosis in Labor (FETAL) Framework: An In-Silico Evaluation of a
1Obstetrics and Gynaecology, Lady Davis Institute (LDI), Jewish General Hospital, Montreal, CAN.
Cardiotocography (CTG) is central to intrapartum fetal surveillance, yet visual interpretation remains limited by poor reproducibility and low specificity for fetal acidemia. Spectral analysis may retain clinically relevant information that is lost when fetal heart-rate variability is reduced to a single amplitude measure or a broad frequency-band ratio, while also providing an interpretable input for clinical decision support. We therefore developed the Fourier Evaluation of Tracings and Acidosis in Labor (FETAL) framework and used Monte Carlo simulation to determine whether its proposed spectral feature could be identified before clinical data collection. In each replicate, a synthetic reference template was constructed from the median normalized power spectrum of 30 independently generated reference-like fetal heart-rate signals. Twelve replicates were performed across five prespecified scenarios: null, weak, moderate, and strong spectral shifts, as well as a moderate shift under artifact stress. Each scenario included 30 reference signals and 60 independent test signals. Ten-minute recordings sampled at 4 Hz incorporated drift, autocorrelated noise, decelerations, missingness, and maternal heart-rate contamination. Power spectra estimated using Welch's method were compared with the reference template using Jensen-Shannon divergence, while deviations in standard deviation and low- to high-frequency ratio served as scalar comparators. Median area under the receiver operating characteristic curve (AUROC) for spectral divergence was 0.504 in the null scenario, 0.466 with weak shifts, 0.778 with moderate shifts, 0.917 with strong shifts, and 0.714 with moderate shifts under artifact stress. The scalar comparators remained near chance across all scenarios. These findings indicate that the FETAL spectral-divergence score can detect graded within-band spectral shifts, remains appropriately uninformative when no usable spectral difference is present, and shows the expected decline in performance under artifact stress. The simulation therefore establishes a locked analytic pipeline, with patient-level validation against umbilical-artery blood gases representing the decisive next step toward clinical implementation.
Cardiotocography (CTG) is central to intrapartum fetal surveillance, yet visual interpretation remains limited by poor reproducibility and low specificity for fetal acidemia. Spectral analysis may retain clinically relevant information that is lost when fetal heart-rate variability is reduced to a single amplitude measure or a broad frequency-band ratio, while also providing an interpretable input for clinical decision support. We therefore developed the Fourier Evaluation of Tracings and Acidosis in Labor (FETAL) framework and used Monte Carlo simulation to determine whether its proposed spectral feature could be identified before clinical data collection. In each replicate, a synthetic reference template was constructed from the median normalized power spectrum of 30 independently generated reference-like fetal heart-rate signals. Twelve replicates were performed across five prespecified scenarios: null, weak, moderate, and strong spectral shifts, as well as a moderate shift under artifact stress. Each scenario included 30 reference signals and 60 independent test signals. Ten-minute recordings sampled at 4 Hz incorporated drift, autocorrelated noise, decelerations, missingness, and maternal heart-rate contamination. Power spectra estimated using Welch's method were compared with the reference template using Jensen-Shannon divergence, while deviations in standard deviation and low- to high-frequency ratio served as scalar comparators. Median area under the receiver operating characteristic curve (AUROC) for spectral divergence was 0.504 in the null scenario, 0.466 with weak shifts, 0.778 with moderate shifts, 0.917 with strong shifts, and 0.714 with moderate shifts under artifact stress. The scalar comparators remained near chance across all scenarios. These findings indicate that the FETAL spectral-divergence score can detect graded within-band spectral shifts, remains appropriately uninformative when no usable spectral difference is present, and shows the expected decline in performance under artifact stress. The simulation therefore establishes a locked analytic pipeline, with patient-level validation against umbilical-artery blood gases representing the decisive next step toward clinical implementation.
