Please use this identifier to cite or link to this item:
Title: A convolutional net for medical records.
Epworth Authors: Wickramasinghe, Nilmini
Other Authors: Nguyen, Phuoc
Truyen, Tran
Venkatesh, Svetha
Keywords: Feature Engineering
Predictive Systems
Electronic Medical Records
End-to-End Deep Learning Syatem
Deep Record
Unplanned Readmission
Chair of Health Informatics Management, Epworth HealthCare, Victoria, Australia
Issue Date: 2017
Publisher: Engineering in Medicine and Biology Society
Citation: IEEE journal of biomedical and health informatics 21.1 (2017): 22-30.
Abstract: Feature engineering remains a major bottleneck when creating predictive systems from electronic medical records. At present, an important missing element is detecting predictive regular clinical motifs from irregular episodic records. We present Deepr (short for Deep record), a new end-to-end deep learning system that learns to extract features from medical records and predicts future risk automatically. Deepr transforms a record into a sequence of discrete elements separated by coded time gaps and hospital transfers. On top of the sequence is a convolutional neural net that detects and combines predictive local clinical motifs to stratify the risk. Deepr permits transparent inspection and visualization of its inner working. We validate Deepr on hospital data to predict unplanned readmission after discharge. Deepr achieves superior accuracy compared to traditional techniques, detects meaningful clinical motifs, and uncovers the underlying structure of the disease and intervention space.
DOI: 10.1109/JBHI.2016.2633963
ISSN: 2168-2194
Journal Title: IEEE Journal of Biomedical and Health Informatics
Type: Journal Article
Affiliated Organisations: Centre for Pattern Recognition and Data Analytics, Faculty of Science and Technology, Deakin University, Australia
Appears in Collections:Health Informatics

Files in This Item:
File Description SizeFormat  
deepr.pdf1.29 MBAdobe PDFView/Open

Items in Epworth are protected by copyright, with all rights reserved, unless otherwise indicated.