RESEARCH
Ontology Driven Temporal Event Annotator mHealth Application Framework
We present an mobile application framework to facilitate the collection of gold standard temporal event annotations. These data will enable training and evaluation of machine learning algorithms for predicting events of clinical significance. Recording of such data using pen and paper can prove to be tedious and error-prone due to the variation in the types of events and the frequency of occurrence. To address this problem, we developed an mHealth application framework that presents an intuitive and configurable user interface for annotating a timeline with events.
Key Takeaways
01. Context
Recent advances in machine learning promise to address many of the most important challenges within healthcare. These advances have been enabled by three elements: algorithmic improvements, increased access to high-performance computing, and dramatic increases in the availability of labeled data from which to learn. This paper focuses on the latter element, i.e. facilitating the rapid collection of large annotated gold-standard datasets for the training and evaluation of supervised machine learning systems.
02. Problem
There is a need for an analogous framework to facilitate the collection of gold standard temporal event annotation data in the context of developing effective systems for predicting temporal events, such as real-time detection of clinical events from physiologic signals. It is important to note that the reusability of most apps for other domains is limited due to tight binding of Application UI to the domain.
03. Solution
We present the Temporal Event Annotator (TEA) framework that aids the process of annotation of clinical events that can be used as gold standard labels for a time series dataset gathered from medical instruments and sensors in a clinical study. The TEA Framework provides a user-centric method to produce a tailored and native mobile application (TEA Fabric) for data collection alongside a complimentary web management client (TEA Central) and a secure API service for seamless data handling and persistence.
04. Impact
This paper discusses the value, design, and development of a real-time Temporal Event Annotator mHealth Application Framework with a dynamic user interface under clinical applications. The TEA Framework has been used to successfully deploy event annotation apps for three studies namely NICU Patient Monitoring, SAANS estimation during VR therapy, and the quantification of patient experience during emergency transport. The TEA Fabric was demonstrated to aid the process of gathering reliable gold-standard temporal event annotations for all the studies.