myAURA: Personalized health library for epilepsy management via knowledge graph sparsification and visualization (2405.05229v2)

Abstract: Objective: We report the development of the patient-centered myAURA application and suite of methods designed to aid epilepsy patients, caregivers, and researchers in making decisions about care and self-management. Materials and Methods: myAURA rests on the federation of an unprecedented collection of heterogeneous data resources relevant to epilepsy, such as biomedical databases, social media, and electronic health records. A generalizable, open-source methodology was developed to compute a multi-layer knowledge graph linking all this heterogeneous data via the terms of a human-centered biomedical dictionary. Results: The power of the approach is first exemplified in the study of the drug-drug interaction phenomenon. Furthermore, we employ a novel network sparsification methodology using the metric backbone of weighted graphs, which reveals the most important edges for inference, recommendation, and visualization, such as pharmacology factors patients discuss on social media. The network sparsification approach also allows us to extract focused digital cohorts from social media whose discourse is more relevant to epilepsy or other biomedical problems. Finally, we present our patient-centered design and pilot-testing of myAURA, including its user interface, based on focus groups and other stakeholder input. Discussion: The ability to search and explore myAURA's heterogeneous data sources via a sparsified multi-layer knowledge graph, as well as the combination of those layers in a single map, are useful features for integrating relevant information for epilepsy. Conclusion: Our stakeholder-driven, scalable approach to integrate traditional and non-traditional data sources, enables biomedical discovery and data-powered patient self-management in epilepsy, and is generalizable to other chronic conditions.

• The paper presents myAURA's multi-layer knowledge graph, integrating diverse data sources for tailored epilepsy management.
• It introduces a network sparsification method that extracts the metric backbone to highlight key clinical and social interactions.
• The findings enhance drug interaction insights and patient care, offering a model for personalized health management in chronic conditions.

Understanding myAURA: A Personal Health Library for Epilepsy Management

Introduction to myAURA

myAURA is an innovative application designed to support epilepsy patients, caregivers, and researchers by integrating diverse health-related data sources into a user-friendly platform. At its core, myAURA encompasses a multi-layer knowledge graph that leverages data from biomedical databases, electronic health records (EHRs), social media platforms, and more, all structured around a human-centered biomedical dictionary. The intent is to provide tailored health information and tools for managing epilepsy more effectively.

Data Integration and Knowledge Graph Construction

How Data is Federated and Processed

myAURA integrates information from a variety of sources:

• Social Media and Community Websites: Platforms like Instagram, Twitter, Reddit, and specialized forums like those of the Epilepsy Foundation of America (EFA) provide real-world data on how patients discuss and manage epilepsy.
• Biomedical and Patient Data: This includes electronic health records and clinical databases that offer insights into drug interactions and treatment outcomes, providing a more clinical perspective on epilepsy management.

Building the Epilepsy Knowledge Graph

The construction of myAURA’s knowledge graph begins by extracting relevant terms from the vast amount of text data these sources generate. With these terms, a network is built where nodes represent terms from the dictionary, and edges represent the co-occurrence of these terms in the textual data. This graph not only illustrates the complex relationships between different aspects of epilepsy care but also provides a structural representation that can be manipulated and analyzed computationally to draw meaningful insights.

Innovative Methods in Network Analysis

Network Sparsification - The Metric Backbone

One of the standout methodologies developed for myAURA is the network sparsification technique that extracts the 'metric backbone' of the knowledge graph. This backbone represents the most crucial and relevant connections between terms, reducing the complexity of the network and focusing on the most significant interactions. This sparsification helps in navigating the graph more effectively, emphasizing important relationships without the noise of less relevant data.

Implications and Applications

The sparsified knowledge graph not only enhances our understanding of drug interactions and patient concerns but also allows for more effective recommendations and visual interpretations of complex data. This can aid patients and healthcare providers in making informed decisions about treatment options based on a comprehensive view of available data.

Future Prospects and Theoretical Contributions

Looking forward, the methodologies and insights from the myAURA project have the potential to be generalized for other chronic conditions, offering a template for integrating heterogeneous data sources into personalized health management tools. On a theoretical level, the project contributes to our understanding of knowledge graphs and network analysis, particularly in how data sparsification can be leveraged to highlight significant information in large datasets.

Conclusion

The myAURA project represents a significant step forward in using integrated data analysis to support patient self-management and medical research in epilepsy. By creating a scalable, stakeholder-driven approach to data integration and application development, myAURA not only serves as a practical tool for epilepsy management but also sets a precedent for future developments in personalized healthcare technology.