EMBC Special Issue: Modeling Early Oxygenation Trajectory in PARDS from High-Frequency Mechanical Ventilation Signals Using Deep Sequence Architectures.

Abstract

Pediatric acute respiratory distress syndrome (PARDS) causes substantial morbidity in the pediatric intensive care unit. We investigated whether high-frequency ventilator waveforms collected during the first hour of invasive ventilation contain predictive information about subsequent oxygenation trajectory. Mechanically ventilated pediatric patients were analyzed using engineered statistical features and learned representations derived from high-frequency ventilator signals, including breath-by-breath flow waveforms from bedside ventilators. Engineered statistical features included early oxygenation indices and ventilator variables from time- and frequency-domain analyses, while learned representations were obtained from 1D-CNN embeddings with PCA-based dimensionality reduction. Two prediction tasks (12-hour OSI regression and classification of mild vs. moderate-or-higher impairment, OSI $\geq 7.5$) were evaluated using five sequence architectures (RNN, LSTM, GRU, Transformer, Mamba) across nine feature configurations, with repeated cross-validation and evaluation on held-out test and temporally separated validation cohorts. For classification, the best cross-validated AUROC was $0.819 \pm 0.030$. Performance remained consistent across independent cohorts, with AUROC values of 0.787-0.823 on the test cohort and 0.808-0.832 on the temporal validation cohort, and Brier scores $\sim$0.12-0.15. For regression, the best cross-validated RMSE was $2.45 \pm 0.81$. Test RMSE ranged from 2.99 to 3.48 OSI units and validation RMSE from 2.60 to 2.85. High-frequency ventilator waveforms acquired during the first hour of mechanical ventilation contain measurable information about short-horizon oxygenation trajectory in PARDS. These findings demonstrate the feasibility of modeling early oxygenation trajectory using routinely available ventilator waveform data and support further prospective and multi-institutional validation prior to clinical translation.