BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Predicting Mortality and Length of Stay with Patient Graph Represe
 ntation Learning - Emma Rocheteau
DTSTART:20201020T121500Z
DTEND:20201020T131500Z
UID:TALK152929@talks.cam.ac.uk
CONTACT:Mateja Jamnik
DESCRIPTION:"Join us on Zoom":https://zoom.us/j/99166955895?pwd=SzI0M3pMVE
 kvNmw3Q0dqNDVRalZvdz09\n\nRecent work on predicting patient outcomes in th
 e Intensive Care Unit (ICU) has focused heavily on the physiological time 
 series data\, largely ignoring sparse data such as diagnoses and medicatio
 ns. When they are included\, they are usually concatenated in the late sta
 ges of a model\, which may struggle to learn from rarer disease patterns. 
 Instead\, we propose a strategy to exploit diagnoses as relational informa
 tion by connecting similar patients in a graph. To this end\, we propose L
 STM-GNN for patient outcome prediction tasks: a hybrid model combining Lon
 g Short-Term Memory networks (LSTMs) for extracting temporal features and 
 Graph Neural Networks (GNNs) for extracting the patient neighbourhood info
 rmation. We demonstrate that LSTM-GNNs outperform the LSTM-only baseline o
 n length of stay prediction tasks on the eICU database. More generally\, o
 ur results indicate that exploiting information from neighbouring patient 
 cases using graph neural networks is a promising research direction\, yiel
 ding tangible returns in supervised learning performance on EHRs.
LOCATION:Zoom
END:VEVENT
END:VCALENDAR