Obtaining interpretable visualization to get meaningful insights of multivariate time series in the intensive care units is very important. A low dimensional representation in which patients' pathology trajectories are apparent and relevant health features are highlighted could benefit clinicians more. On eICU dataset it was demonstrated that, T-DPSOM gives interpretable visualisations of patient state trajectory and uncertainty estimation. Initially a novel way of fitting SOMs with probabilistic cluster assignments, named as PSOM is used. Then a deep architecture is proposed, Deep Probabilistic SOM, for training of a VAE and a PSOM together to obtain interpretable discrete representations. This model is further extended to support time series data, giving the temporal DPSOM (T-DPSOM). T-DPSOM can be used cluster patients into different sub-phenotypes and can gain better insights for disease patterns and individual patient health state. For that purpose eICU dataset is used, which contains multivariate medical time series from the ICU. The variables include vital signs and lab measurements. Predictions are based on APACHE score and heatmaps are used for the purpose of visualization.
Input variables : Multivariate Time Series (Lab measurements, Vital signs)
Output Variables : Patients State (APACHE Score)
Business | : | Population at High Risk of Disease | Risk by Geography | Risk by Demographics | Risk by Clinical Parameters | Optimized Hospital Resource Utilization | Decreased Cost of Care | Decreased Patient Visits |
Infrastructure | : | Log Bytes | Logging/User/IAMPolicy | Logging/User/VPN | CPU Utilization | Memory Usage | Error Count | Prediction Count | Prediction Latencies | Private Endpoint Prediction Latencies | Private Endpoint Response Count |
Visit Model : github.com
Additional links : mds.inf.ethz.ch
Model Category | : | Public |
Date Published | : | March, 2021 |
Healthcare Domain | : | Provider |
Code | : | github.com |
Health Risk Management |
Health Risk Prediction |