Fetal cardiovascular decompensation during labor predicted from the individual heart rate: a prospective study in fetal sheep near term and the impact of low sampling rate

TL;DR

This study introduces a real-time machine learning algorithm that predicts fetal cardiovascular decompensation during labor with high accuracy, even with noisy data, and highlights the importance of sampling rate for optimal performance.

Contribution

The paper presents a novel individualized fetal heart rate analysis algorithm capable of early prediction of cardiovascular decompensation during labor.

Findings

Algorithm achieves 92% sensitivity on high-quality data.

Performance drops to 67% sensitivity with low sampling rate data.

Real-time prediction can potentially improve fetal monitoring during labor.

Abstract

We present a novel computerized fetal heart rate intrapartum algorithm for early and individualized prediction of fetal cardiovascular decompensation, a key event in the causal chain leading to brain injury. This real-time machine learning algorithm performs well on noisy fetal heart rate data and requires ~2 hours to train on the individual fetal heart rate tracings in the first stage of labor; once trained, the algorithm predicts the event of fetal cardiovascular decompensation with 92% sensitivity. We show that the algorithm's performance suffers reducing sensitivity to 67% when the fetal heart rate is acquired at the sampling rate of 4 Hz used in ultrasound cardiotocographic monitors compared to the electrocardiogram(ECG)-derived signals as can be acquired from maternal abdominal ECG.