New CT-HEG graph schema improves ICU mortality prediction

Researchers have developed a new schema called CT-HEG (Continuous-Time Heterogeneous EHR Graph) to model complex clinical data for predicting in-hospital mortality in ICUs. This schema represents ICU stays as timestamped graphs, incorporating node types for visits, vitals, and lab events, with edge attributes capturing timing and value. When implemented as the CHIRP-Net model, it achieved a mean AUROC of 0.8449 on the MIMIC-IV v3.1 dataset, outperforming several baseline models. An ablation study indicated that bidirectional connectivity was essential for the model's functionality, and time-attentive edge features significantly improved performance. AI

IMPACT This new graph schema and model architecture could enhance the accuracy of clinical predictions in healthcare settings.

RANK_REASON The cluster contains an academic paper detailing a new model architecture and schema for a specific task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• CHIRP-Net
• CT-HEG
• GATv2Conv
• logistic regression model
• MIMIC-IV v3.1
• Mohammad Nasir Uddin
• Transformer++

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →