Using Knowledge Graphs and Neural Networks for Enhanced Medical Research and Discovery (AIMultimediaLab, ContinualBot, 2024)
Info
- Organizer Universitatea de Medicină și Farmacie "Carol Davila"
- Location Palace of Parliament, Bucharest, Romania
- Project type Doctoral Research Poster
- Date 25th of Oct 2024
- Participants ~30
- My Role PhD Researcher, Speaker
- Topics Understanding Vast Medical Literature, Data-Driven Decisions, Novel Insights, Natural Language, Research
- Keywords Artificial Intelligence (AI), Neural Networks, Knowledge Graphs, Medical Research, Large Language Models (LLMs), Transformer Architecture, Retrieval-Augmented Generation (RAG), Data Processing in Medicine, Graph Theory, Medical Knowledge Representation
- Skills developed Research, GraphRAG
- Presentation Slides URL
Description
Advancements in technology and software have profoundly supported and transformed various scientific domains, including medical research, by providing innovative methodologies for expanding scientific knowledge. A significant aspect of this progress involves addressing the challenges associated with processing and analyzing extensive volumes of medical data, a task that has historically been complex and resource-intensive for researchers. Neural networks, particularly those employing transformer architectures, have revolutionised data processing capabilities, facilitating the development of large language models (LLMs) capable of managing substantial amounts of information effectively.
In this study, we investigate the integration of Knowledge Graphs and Neural Networks to improve the efficiency and accuracy of medical research analysis. By utilizing vector and graph databases, we construct Retrieval-Augmented Generation (RAG) systems that reveal previously unrecognized relationships and patterns within medical studies. These systems enable the simultaneous processing of up to one million tokens, equivalent to analyzing the entire Romanian Explanatory Dictionary (DEX). This capability represents a significant advancement in synthesizing and interrogating extensive research data sets.
Our poster presents a case study wherein a collection of scientific medical research papers is transformed into a comprehensive knowledge base using graph theory principles. This structured knowledge base provides a robust contextual framework, enabling AI systems to systematically read, interpret, and interact with the research literature, thereby uncovering novel insights and expediting the pace of medical discoveries. This approach holds substantial potential for improving diagnostic processes and advancing the overall field of medical science through enhanced data-driven exploration
About the Project
This project forms a core part of my PhD research and reflects ongoing work conducted in collaboration within our research team at AIMultimediaLab. We are exploring the intersection of artificial intelligence and medical research, focusing on how advanced systems like Neural Networks and Knowledge Graphs can revolutionize the way vast amounts of medical data are processed, analyzed, and understood.
Our system, built on a Retrieval-Augmented Generation (RAG) framework, combines deep learning and knowledge graph technologies to enable a deeper, more contextual understanding of medical data. By integrating these advanced computational techniques, our system offers unprecedented support to medical professionals and researchers, enhancing the discovery of novel relationships and accelerating the diagnosis and treatment of complex medical conditions.
A central case study of this work focuses on Burning Mouth Syndrome, where the system has demonstrated the ability to analyze proprietary data to uncover new correlations between patient profiles and the syndrome's presentation. The poster presentation highlights both the technical framework of the system and practical examples of its application in the medical field.
My Contribution
This project represents a significant personal and academic milestone. As part of my PhD research, I led the development of the AI-assisted medical research platform, overseeing both the integration of the Neural Networks and Knowledge Graphs into the RAG system and the clinical application of this technology. The platform not only assists researchers in processing vast datasets but also ensures that language barriers are eliminated through its multilingual query-based interface.
This project aims to enhance the efficiency of medical research by enabling doctors and researchers to ask complex questions in natural language, retrieving answers from both structured and unstructured data stored in proprietary databases. The ultimate goal is to reduce the cognitive load on medical practitioners, enabling faster, more informed decision-making.
Grateful Acknowledgment
I would like to express my heartfelt gratitude to Dr. Cosmin Dugan, who has supported and guided this work. His invitation to present this project at UMF Carol Davila is an honor, and I am deeply appreciative of his collaboration.
Stay tuned for further updates from the AIMultimediaLab as we continue to push the boundaries of what AI can achieve in the medical field.
