RESEARCH

Time Series Representation Learning 📈

Irregularly sampled time series in EHR

Graph-based embedding of missing information
Learning balanced representation for causal relationship modeling
Disease detection and counterfactual outcome simulation for treatment effect evaluation

Temporal predictive soft clustering

Predictive soft clustering for time series using variational inference of GMM
Continuous phenotyping of clinical deterioration
Embedding space for counterfactual simulation of treatments

Contrastive learning for EHR

Supervised contrastive learning
Deep representation learning for EHR
Prognostic monitoring system for shock in ICU patients

Multi-task learning for continuous risk prediction

Multi-task learning prediction model
Model optimization for MTL
Continuous early warning system for acute kidney injury

EHR Multimodal Learning 📊🧬📄

Multimodal fusion of EHR

Robust representation learning from incomplete multimodal clinical data
Cross-modal temporal alignment
Online risk monitoring framework for pulmonary diseases

Wearable IoT-enabled health monitoring system

Deep representation learning for EHR
Time series clustering of cardiac rehabilitation recovery pattern
Wearable monitoring device

Reinforcement Learning 🖥️

Frictional off-policy RL

Physical anaology for representing extrapolation error
Expanded and improved constraint with replay buffer
Contrastive variational autoencoder using orthogonal action space

Safe and reasonable medical AI

Clinician-informed RL for shock treatment
Multi-agent RL framework for enhancing clinical stability
Treatment decision system integrating clinical reasoning

Medical dead-ends for safe RL

Safe reinforcement learning
Multi-agent deep reinforcement learning
Identifying high risk states and management of nephrotoxic drugs

Multimodal RL for embedding networks

Deep reinforcement learning
Multimodal data integration
Medication recommendation for Parkinsons' disease

Data-driven identification of drug-ADR relationships

Deep reinforcement learning
Network analysis and graph theory
Identifying nephrotoxic drug-drug interactions

Language Models for EHR 💬

LLM-based representation learning for time series

Inference of missing vital sign values using a pre-trained LLM
Lab result embedding based on temporal similarity between observed time and overall time
Deriving feature importance through variable self-attention

EHR sequence language modeling

Pretrained language model
Long-document transformer for EHR
EHR sequence representation for enhanced predictive modeling

Time Series Anomaly Detection 🔎

Sequential anomaly detection for clinical deterioration

Time series anomaly detection
Early warning system for clinical deterioration
Detection of unexpected ICU admissions from ED

🏥 Medical Partners 🧑‍🤝‍🧑

Page updated

Report abuse