Predicting Fetal Outcomes from Cardiotocography Signals Using a Supervised Variational Autoencoder

2025-09-08

AI Summary

To address the limited interpretability of deep learning models in cardiotocography (CTG) signal classification, this paper proposes a supervised variational autoencoder (SVAE) that jointly optimizes signal reconstruction and fetal outcome prediction. By incorporating KL-divergence constraints and total correlation regularization, the latent space is explicitly disentangled to enhance clinical interpretability. We introduce a novel quantitative interpretability assessment framework integrating latent traversal, R² analysis, and unsupervised component decomposition. Results reveal that baseline fetal heart rate features are robustly and distinctly encoded in the latent representation, whereas variability metrics exhibit weaker encoding. The model achieves AUROC scores of 0.752 (segment-level) and 0.779 (full-CTG-level) for fetal outcome prediction. Notably, relaxing the total correlation constraint simultaneously improves both reconstruction fidelity and classification performance. This work establishes a new paradigm for clinically trustworthy, interpretable CTG intelligent interpretation.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Interpretability, Explainability, and Transparency
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Objective: To develop and interpret a supervised variational autoencoder (VAE) model for classifying cardiotocography (CTG) signals based on pregnancy outcomes, addressing interpretability limits of current deep learning approaches. Methods: The OxMat CTG dataset was used to train a VAE on five-minute fetal heart rate (FHR) segments, labeled with postnatal outcomes. The model was optimised for signal reconstruction and outcome prediction, incorporating Kullback-Leibler divergence and total correlation (TC) constraints to structure the latent space. Performance was evaluated using area under the receiver operating characteristic curve (AUROC) and mean squared error (MSE). Interpretability was assessed using coefficient of determination, latent traversals and unsupervised component analyses. Results: The model achieved an AUROC of 0.752 at the segment level and 0.779 at the CTG level, where predicted scores were aggregated. Relaxing TC constraints improved both reconstruction and classification. Latent analysis showed that baseline-related features (e.g., FHR baseline, baseline shift) were well represented and aligned with model scores, while metrics like short- and long-term variability were less strongly encoded. Traversals revealed clear signal changes for baseline features, while other properties were entangled or subtle. Unsupervised decompositions corroborated these patterns. Findings: This work demonstrates that supervised VAEs can achieve competitive fetal outcome prediction while partially encoding clinically meaningful CTG features. The irregular, multi-timescale nature of FHR signals poses challenges for disentangling physiological components, distinguishing CTG from more periodic signals such as ECG. Although full interpretability was not achieved, the model supports clinically useful outcome prediction and provides a basis for future interpretable, generative models.

Problem

• Classifying cardiotocography signals for fetal outcome prediction
• Addressing interpretability limits in deep learning CTG analysis
• Developing supervised variational autoencoder for meaningful feature encoding

Innovation

• Supervised variational autoencoder for CTG classification
• KL divergence and total correlation constraints applied
• Latent space analysis for clinical feature interpretation