Temporal Data Mining for Detecting Adverse Events in Healthcare

  • Funding Organisation: Swedish Research Council (Vetenskapsrådet)
  • Type of Funding: Starting Grant (Etableringsbidrag)
  • Duration: 2017-2020
  • Project Leader: Panagiotis Papapetrou

Abstract: This project aims to develop a generic framework that will advance the current state-of-the-art in the areas of temporal data mining and healthcare informatics. The framework will comprise novel methods and tools for searching and mining massive and heterogeneous data sources of temporal nature. The focus application domain will be healthcare and more particularly the detection of adverse events (AEs) in electronic healthcare records (EHRs). The central properties of the proposed framework will include the ability to: i) efficiently search and index large heterogeneous data sources, ii) effectively exploit and learn from the inherent temporal nature of these data sources, iii) provide search results and predictions with confidence and statistical guarantees. The developed framework will be used for implementing a demonstrator, which will provide clinical practitioners with support for recognizing, understanding, and preventing AEs in healthcare by exploiting EHRs. The project is expected to produce foundational contributions to the areas of temporal data mining and healthcare informatics. The developed algorithms will constitute the building blocks for future indexing and predictive modeling frameworks in temporal data spaces. More importantly, the produced methods and tools are expected to be employed for automated knowledge discovery and information extraction from EHR data sources of complex and heterogeneous nature.

Main Objectives: The project has three concrete objectives each constituting a separate Work Package (WP):

  • WP1: Searching and indexing temporal data spaces
  • WP2: Predictive modeling in complex temporal data sources
  • WP3: Validation on adverse event detection in EHRs