CERVINet-DELTA

# CERVINet-DELTA: Temporal Multimodal Modeling for Labor Outcome Prediction This repository contains the implementation of **CERVINet-DELTA**, a unified framework for dynamic, explainable, and personalized prediction of labor induction outcomes using multimodal clinical data. The system models both latent temporal dynamics and actionable policy optimization to assist obstetric decision-making under uncertainty.  --- ## 📌 Key Features - **CERVINet**: A generative latent-state model with structured encoders and GRU-based transition dynamics.- **DELTA**: A decision-theoretic strategy layer for treatment planning and real-time risk stratification.- **Multimodal Inputs**: Supports structured patient data, time series (e.g., uterine activity), and image embeddings.- **Uncertainty-aware**: Variational inference with personalized outcome distributions.- **Explainable AI**: Saliency analysis and utility gradients for interpretability. --- ## 📂 Project Structure