Cardea: An Open Automated Machine Learning Framework for Electronic Health Records
This addresses the problem of repetitive and inefficient research workflows for researchers in healthcare machine learning, though it is incremental as it builds on existing AUTOML frameworks and standards.
The authors tackled the lack of a trusted software framework for deep learning on electronic health records by proposing Cardea, an open-source automated machine learning system that achieved human-competitive performance on 5 prediction tasks using MIMIC-III and Kaggle datasets.
An estimated 180 papers focusing on deep learning and EHR were published between 2010 and 2018. Despite the common workflow structure appearing in these publications, no trusted and verified software framework exists, forcing researchers to arduously repeat previous work. In this paper, we propose Cardea, an extensible open-source automated machine learning framework encapsulating common prediction problems in the health domain and allows users to build predictive models with their own data. This system relies on two components: Fast Healthcare Interoperability Resources (FHIR) -- a standardized data structure for electronic health systems -- and several AUTOML frameworks for automated feature engineering, model selection, and tuning. We augment these components with an adaptive data assembler and comprehensive data- and model- auditing capabilities. We demonstrate our framework via 5 prediction tasks on MIMIC-III and Kaggle datasets, which highlight Cardea's human competitiveness, flexibility in problem definition, extensive feature generation capability, adaptable automatic data assembler, and its usability.