Deep Learning with Information Fusion and Model Interpretation for Health Monitoring of Fetus based on Long-term Prenatal Electronic Fetal Heart Rate Monitoring Data

TL;DR

The paper introduces LARA, a CNN-based system with information fusion for automatically analyzing long-term antepartum fetal heart rate data. By producing a time-resolved Risk Distribution Map (RDM) and an overall Risk Index (RI), LARA enables both short-term risk classification and long-term outcome association analyses, while providing multiple interpretability angles via output scores, Grad-CAM attention, and deep-feature analysis. The study demonstrates that LARA achieves strong short-term classification performance (AUC up to 0.872 after resampling) and reveals a significant link between RI and adverse outcomes such as Small for Gestational Age (SGA), with qualitative insights into model behavior and feature relevance. These findings suggest clinical value for long-term FHR monitoring, offering a path toward automated, interpretable, home-based fetal surveillance and reduced manual burden on clinicians.

Abstract

Long-term fetal heart rate (FHR) monitoring during the antepartum period, increasingly popularized by electronic FHR monitoring, represents a growing approach in FHR monitoring. This kind of continuous monitoring, in contrast to the short-term one, collects an extended period of fetal heart data. This offers a more comprehensive understanding of fetus's conditions. However, the interpretation of long-term antenatal fetal heart monitoring is still in its early stages, lacking corresponding clinical standards. Furthermore, the substantial amount of data generated by continuous monitoring imposes a significant burden on clinical work when analyzed manually. To address above challenges, this study develops an automatic analysis system named LARA (Long-term Antepartum Risk Analysis system) for continuous FHR monitoring, combining deep learning and information fusion methods. LARA's core is a well-established convolutional neural network (CNN) model. It processes long-term FHR data as input and generates a Risk Distribution Map (RDM) and Risk Index (RI) as the analysis results. We evaluate LARA on inner test dataset, the performance metrics are as follows: AUC 0.872, accuracy 0.816, specificity 0.811, sensitivity 0.806, precision 0.271, and F1 score 0.415. In our study, we observe that long-term FHR monitoring data with higher RI is more likely to result in adverse outcomes (p=0.0021). In conclusion, this study introduces LARA, the first automated analysis system for long-term FHR monitoring, initiating the further explorations into its clinical value in the future.

Figures (8)

• Figure 1: Construction of prediction model This figure illustrates the procedural steps involved in the construction of our prediction model. cov1d: Convolution layer designed for 1-Dimensional input. BN: Batch Normalization layer. Swish: An activation function. SE: Squeeze-and-Excitation layer. GAP: Global Average Pooling layer. FC: Fully Connected layer.
• Figure 2: Information Fusion process This figure delineates the stages involved in the process of information fusion. Cam$\_$i is calculated using the Grad-CAM algorithm. CNN: Convolutional Neural Network. mRI: risk index for minute-unit.
• Figure 3: Flow chart of normal and abnormal group recruitment
• Figure 4: Model performance (a) Receiver operating characteristic (ROC) curve of the convolutional neural network (CNN) model for the test dataset. (b) Confusion matrix depicting the classification outcomes of the model, in conjunction with the threshold applied based on the Youden Index. (c) (d) Distribution of predicted scores among manually labeled five groups and two groups, illustrated through the density plot and boxplot, respectively. In (c), the five groups pertains to the categorization based on manually assigned scores according to the Scoring rubric outlined in Table 1. In (d), the classification criterion for normal is established as scores greater than 3, whereas abnormal is defined as scores equal to or less than 3.
• Figure 5: RI validate for fetal outcomes Distribution of Risk Index (RI) for long-term FHR across various fetal outcomes. Criteria 1 delineates a positive outcome as the manifestation of either maternal superimposed preeclampsia (SPE), the coexistence of chronic hypertension complicated with preeclampsia (PE), or the fetus falling below the 3rd percentile for Small for Gestational Age (SGA). For criteria 2, positive is defined as fetal brain disorder, including intraventricular hemorrhage, cerebral white matter softening, enhanced echogenicity in the cerebral white matter, and cerebral white matter loss. In the combined assessment of Criteria 1 and 2 (denoted as Criteria 1 + 2), a positive determination is established if either Criteria 1 or Criteria 2 is met.