A New Deep State-Space Analysis Framework for Patient Latent State Estimation and Classification from EHR Time Series Data
21 July 2023
Developed deep learning model to estimate latent disease states from EHRs over time
Applied model to 12,695 cancer patients undergoing chemotherapy
Revealed latent states correlating with prognosis and mortality
Visualized interpretable transitions in disease state over time
Identified key clinical factors during critical state transitions
Deep learning model estimates cancer patient health from records
Researchers developed a deep learning model to analyze electronic health records over time and estimate latent disease states in cancer patients. This could help understand disease progression and guide treatment. The model was applied to records of 12,695 cancer patients, revealing interpretable latent states related to prognosis. Visualizing state transitions showed progression from stable to unstable to death. Analyzing by drug identified important clinical factors during critical transitions.
No comments yet, be the first to start the conversation...
Sign up to comment on this paper