The abstract is read at the publisher; the summary is JClub's.
Marked DeepHit predicts imminent delivery within one hour from intrapartum CTG, achieving test-set AUROC of 0.837 at 6 hours and 0.935 at 12 hours into labor.
- Why it matters: Most CTG machine learning assumes a known 30–60 minute window before delivery, which is unrealistic; reframing as time-to-event aims to enable earlier, timing-agnostic fetal distress detection.
- What they did: The study builds deep discrete-time survival and marked point process models to estimate joint risk of imminent delivery and fetal acidosis, using continuous physiological signals with fully observed events and partially observed labels.
- The result: The approach estimates delivery at multiple elapsed-time landmarks and shows competitive near-delivery performance versus end-of-tracing models, outperforming clinical feature-based methods on internal and external validation at a second site.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.