TY - GEN
T1 - Heterogeneous Graph Embeddings of Electronic Health Records Improve Critical Care Disease Predictions
AU - Wanyan, Tingyi
AU - Kang, Martin
AU - Badgeley, Marcus A.
AU - Johnson, Kipp W.
AU - De Freitas, Jessica K.
AU - Chaudhry, Fayzan F.
AU - Vaid, Akhil
AU - Zhao, Shan
AU - Miotto, Riccardo
AU - Nadkarni, Girish N.
AU - Wang, Fei
AU - Rousseau, Justin
AU - Azad, Ariful
AU - Ding, Ying
AU - Glicksberg, Benjamin S.
N1 - Publisher Copyright:
© 2020, Springer Nature Switzerland AG.
PY - 2020
Y1 - 2020
N2 - Electronic Health Record (EHR) data is a rich source for powerful biomedical discovery but it consists of a wide variety of data types that are traditionally difficult to model. Furthermore, many machine learning frameworks that utilize these data for predictive tasks do not fully leverage the inter-connectivity structure and therefore may not be fully optimized. In this work, we propose a relational, deep heterogeneous network learning method that operates on EHR data and addresses these limitations. In this model, we used three different node types: patient, lab, and diagnosis. We show that relational graph learning naturally encodes structured relationships in the EHR and outperforms traditional multilayer perceptron models in the prediction of thousands of diseases. We evaluated our model on EHR data derived from MIMIC-III, a public critical care data set, and show that our model has improved prediction of numerous disease diagnoses.
AB - Electronic Health Record (EHR) data is a rich source for powerful biomedical discovery but it consists of a wide variety of data types that are traditionally difficult to model. Furthermore, many machine learning frameworks that utilize these data for predictive tasks do not fully leverage the inter-connectivity structure and therefore may not be fully optimized. In this work, we propose a relational, deep heterogeneous network learning method that operates on EHR data and addresses these limitations. In this model, we used three different node types: patient, lab, and diagnosis. We show that relational graph learning naturally encodes structured relationships in the EHR and outperforms traditional multilayer perceptron models in the prediction of thousands of diseases. We evaluated our model on EHR data derived from MIMIC-III, a public critical care data set, and show that our model has improved prediction of numerous disease diagnoses.
KW - Electronic health records
KW - Embeddings
KW - Heterogeneous graph learning
KW - Skip-gram model
UR - http://www.scopus.com/inward/record.url?scp=85092245325&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-59137-3_2
DO - 10.1007/978-3-030-59137-3_2
M3 - Conference contribution
AN - SCOPUS:85092245325
SN - 9783030591366
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 14
EP - 25
BT - Artificial Intelligence in Medicine - 18th International Conference on Artificial Intelligence in Medicine, AIME 2020, Proceedings
A2 - Michalowski, Martin
A2 - Moskovitch, Robert
PB - Springer Science and Business Media Deutschland GmbH
T2 - 18th International Conference on Artificial Intelligence in Medicine, AIME 2020
Y2 - 25 August 2020 through 28 August 2020
ER -