Publications & Research

Conference Paper · CCAI 2026

Deep Learning for Stroke Mortality Prediction in eICU: A Dual-Tower Transformer Framework

Zhengrong Jia*, Kwong-Cheong Wong*

The 6th International Conference on Computer Communication and Artificial Intelligence (CCAI), 2026

Abstract

We propose a novel Dual-Tower Transformer (DT-Transformer) for stroke mortality prediction on the multicenter eICU Collaborative Research Database. The decoupled architecture processes categorical demographics and numerical vitals through separate tower pathways, each equipped with Self-Attention, before fusing representations for final prediction. The model achieves an AUPRC of 0.6171 — a 14.4% improvement over the strongest neural baseline. An Adaptive Runtime Safeguard is integrated for inference stability against physiological outliers, and attention map visualizations provide clinical interpretability.

PDF (Preprint) Code (GitHub) Slides ORCID Google Scholar
Key Results
0.6171
AUPRC (std 0.006)
14.4%
vs Strongest Neural Baseline
5-Fold
Cross-Validation
Architecture & Methodology
  • Dual-Tower Design — Separate encoding pathways for categorical (demographics) and numerical (vitals) features
  • Self-Attention Layers — Each tower applies multi-head Self-Attention for intra-modality feature interaction
  • Late Fusion — Tower outputs concatenated before classification head
  • Adaptive Runtime Safeguard — Detects out-of-distribution inputs at inference time for clinical safety
  • Attention Visualization — Heatmaps over input features for clinical interpretability
Python Transformer Self-Attention eICU XGBoost PyTorch Scikit-learn
Dataset & Reproducibility

eICU Collaborative Research Database

  • Multicenter critical care database (200,859 admissions, 208 hospitals)
  • Stroke cohort extracted with ICD-9 codes 430–438
  • Features: demographics, vitals, lab values, GCS scores
  • Access via PhysioNet credentialed access

Code Repository

Open-source implementation available on GitHub. Repository includes:

  • PyTorch model implementation
  • Data preprocessing pipeline
  • 5-fold cross-validation scripts
  • Attention visualization tools
How to Cite

If you find this work useful, please cite:

@inproceedings{jia2026dttransformer,
  title     = {Deep Learning for Stroke Mortality Prediction
               in eICU: A Dual-Tower Transformer Framework},
  author    = {Jia, Zhengrong and Wong, Kwong-Cheong},
  booktitle = {Proceedings of the 6th International Conference on
               Computer Communication and Artificial Intelligence (CCAI)},
  year      = {2026},
  note      = {Accepted. To appear in May 2026}
}
Baseline Comparison
Model AUROC AUPRC
DT-Transformer 0.8848 ± 0.0034 0.6171 ± 0.0058
XGBoost 0.8908 0.6467
Random Forest 0.8806 0.6236
NN (Neural Network) 0.8582 ± 0.0018 0.5394 ± 0.0054
Standard Transformer 0.8457 ± 0.0129 0.5279 ± 0.0195
Standard MLP 0.8534 ± 0.0058 0.5170 ± 0.0081