CERVINet-DELTA: Multimodal Temporal Modeling for Labor Induction Outcomes

# CERVINet: Multimodal Attention-Based Prediction for Dinoprostone-Induced Labor Outcomes > A temporal probabilistic modeling framework for personalized labor induction decision support using multimodal data fusion and attention mechanisms. ## 🌟 Overview This repository contains the implementation of **CERVINet** and **DELTA**, two complementary frameworks designed to improve clinical decision-making in labor induction via: - Multimodal data integration (clinical features, uterine activity, ultrasound)- Temporal latent dynamics modeling- Outcome-aware policy optimization- Explainable attention mechanisms Our models are benchmarked on two real-world datasets:- **Dinoprostone Dataset**- **Multimodal Labor Dataset** ## 🧠 Highlights - 📈 Outperforms state-of-the-art time series predictors like Transformer, Informer, and N-BEATS- 🧩 Modular ablation analysis for key components (latent dynamics, multivariate heads, policy layers)- 🔎 Attention-based interpretability across modalities and time- ⚙️ End-to-end training with uncertainty quantification via variational inference ## 📁 Project Structure