Deep Semi-Supervised Embedded Clustering (DSEC) for Stratification of\n Heart Failure Patients

Determining phenotypes of diseases can have considerable benefits for\nin-hospital patient care and to drug development. The structure of high\ndimensional data sets such as electronic health records are often represented\nthrough an embedding of the data, with clustering methods used to group data of\nsimilar structure. If subgroups are known to exist within data, supervised\nmethods may be used to influence the clusters discovered. We propose to extend\ndeep embedded clustering to a semi-supervised deep embedded clustering\nalgorithm to stratify subgroups through known labels in the data. In this work\nwe apply deep semi-supervised embedded clustering to determine data-driven\npatient subgroups of heart failure from the electronic health records of 4,487\nheart failure and control patients. We find clinically relevant clusters from\nan embedded space derived from heterogeneous data. The proposed algorithm can\npotentially find new undiagnosed subgroups of patients that have different\noutcomes, and, therefore, lead to improved treatments.\n