New contrastive learning method improves cross-national EHR data transfer

Researchers have developed a new pre-training objective called Asymmetric Supervised Contrastive Learning (Asymmetric SupCon) to improve the transferability of Electronic Health Record (EHR) representations across different national healthcare systems. This method focuses on clustering patients with positive outcomes without explicitly grouping negative ones, addressing the heterogeneity in clinical data. Pre-training temporal Transformer encoders on a large Taiwanese EHR dataset and transferring them to U.S. datasets like MIMIC-IV and EHRSHOT demonstrated significant performance improvements, particularly in few-shot learning scenarios for disease prediction. AI

IMPACT This research could enable more effective cross-border collaboration and analysis of medical data, potentially accelerating medical research and improving patient care globally.

RANK_REASON The cluster contains a research paper detailing a new method for machine learning on medical data. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Asymmetric SupCon
• Asymmetric Supervised Contrastive Learning
• EHRSHOT
• EHR
• MIMIC-IV
• NHIRD
• Qingyang Zhang
• transformer

• paper
• model release

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