Uncovering Trajectory and Topological Signatures in Multimodal Pediatric Sleep Embeddings

2026-05-13

AI Summary

This study addresses the limited clinical interpretability of pediatric multimodal sleep embeddings by proposing a novel late-fusion framework that jointly leverages PHATE trajectory coordinates, persistent homology topological summaries, and electronic health records (EHR). By integrating geometric, topological, and clinical signals through both linear and multilayer perceptron (MLP) models, the approach uncovers complementary patterns within embedding sequences. Evaluated under extreme class imbalance, the method significantly enhances model calibration and robustness, achieving substantial improvements in area under the precision-recall curve (AUPRC) across four binary classification tasks—for instance, increasing AUPRC for EEG arousal detection from 0.31 to 0.48—and attaining state-of-the-art calibration performance.

Abstract

While generative models have shown promise in pediatric sleep analysis, the latent structure of their multimodal embeddings remains poorly understood. This work investigates session-wide diagnostic information contained in the sequences of 30-second pediatric PSG epochs embedded by a multimodal masked autoencoder. We test whether augmenting embeddings with PHATE-derived per-epoch coordinates and whole-night movement descriptors, persistent homology summaries of the embedding cloud, and EHR yields task-relevant signals. Simple linear and MLP models, chosen for interpretability rather than state-of-the-art performance, show that geometric, topological, and clinical features each provide complementary gains. For binary predictions, feature importance is task-dependent, and more expressive late-fusion models generally perform better, with AUPRC improving from 0.26 to 0.34 for desaturation, 0.31 to 0.48 for EEG arousal, 0.09 to 0.22 for hypopnea, and 0.05 to 0.14 for apnea. We also report Brier score and Expected Calibration Error, where the full fusion model yields the best calibration across all four binary tasks. Our study reveals that latent geometry/topology and EHR offer complementary, interpretable signals beyond embeddings, improving calibration and robustness under extreme imbalance.

Problem

• multimodal embeddings
• pediatric sleep
• topological signatures
• latent geometry
• clinical prediction

Innovation

• multimodal embeddings
• persistent homology
• PHATE
• EHR integration
• topological data analysis