Large language models and knowledge graphs: solutions for healthcare data management

May, 2024.- A proposal that combines the use of large language models with data graphs for clinical data: this is the innovation proposed in the article Augmented non-hallucinating large language models as medical information curators published in NPJ Digital Medicine, by researchers Stephen Gilbert and Jakob Nikolas Kather, from the Else Kröner Fresenius Center for Digital Health at the Dresden University of Technology, Germany, together with the deputy director of the Millennium Institute for Foundational Research on Data (IMFD), Aidan Hogan, from the Department of Computer Science of the Universidad de Chile.

Medical data presents a major challenge for current data science: it is complex, must necessarily be safeguarded, and there are a number of challenges to managing it efficiently. All of this complexity has driven the search for alternatives that can enable its application and allow this valuable information to be used by healthcare providers, for the benefit of their patients.

Large language models (LLMs) -very large deep learning models trained on huge amounts of data, the foundation of applications such as ChatGPT-, can contribute significantly to better structuring, categorizing, and interpreting medical information, however, they have weaknesses that make their use in an area as critical as health care difficult. The generation of plausible but incorrect information (hallucinations), or the ability to give different answers to the same query, make their use in medicine very complex. On the other hand, knowledge graphs are a way of structuring large amounts of information that can be in various formats or multivariable, through the use of nodes and edges that connect the data.

The researchers from the EKFZ for Digital Health at TU Dresden and the Universidad de Chile propose, in the article Augmented non-hallucinating large language models as medical information curators published, a possible solution to this problem: the combination of LLMs with knowledge graphs (KG). This gives rise to a new form of Retrieval Augmented Generation, which would allow models to be more reliable, robust, and capable of reproducing queries.

The reliable recording of medical information and its exchange between different systems (interoperability) is a major challenge in the health field and is often called the “communication problem” of medicine. Medical ontologies and knowledge graphs (KG) are approaches to solving this problem. Medical ontologies function as dictionaries of medical terms that help categorize and define medical concepts. However, since in human language terms can have different meanings depending on context, these ontologies tend to be ambiguous. The word “cold,” in English, for example, can refer to body temperature, environmental conditions, or a common cold. The same happens in all languages, with differences between different disciplines within health care. The use of acronyms is another major challenge in the field: COLD can also mean “Chronic obstructive lung disease.”

Knowledge graphs (KG) are organized networks that connect different medical concepts and their relationships. For example, the term “COVID-19” in a graph could be connected to “fever” through a link called “has symptom.” The graphs facilitate the understanding and processing of medical information, but face challenges similar to those of medical ontologies.

Combination for structured reasoning

To address these shortcomings, the researchers from Dresden and Santiago, Chile propose combining LLMs with KGs, leveraging their respective strengths. This combination provides structured reasoning and could help reduce model bias and offer more reliable, accurate, and reproducible results. These approaches would be more compatible with regulatory approval pathways than LLMs alone.

“The combination of large language models and knowledge graphs is a way of linking existing medical knowledge with the cognitive capabilities of large language models. We are only at the beginning of a very exciting development,” says Professor Jakob N. Kather, Chair of Clinical Artificial Intelligence at the Dresden University of Technology and oncologist at the Carl Gustav Carus University Hospital in Dresden.

In the research, the authors discuss different approaches to combining LLMs with KGs. They suggest that this could also facilitate the development of robust “digital twins” of patients, in the form of structured individual medical records that enable personalized diagnosis.

“Although regulatory challenges remain, healthcare professionals graduating today can anticipate access to compliant and advanced clinical information summarization tools that were unimaginable just five years ago. Furthermore, approaches that combine large language models with knowledge graphs are more likely to achieve early approval in conservative regulatory pathways,” says Professor Stephen Gilbert, Chair of Regulatory Science of Medical Devices at the Dresden University of Technology.

For Aidan Hogan, “Just like LLMs, KGs have applications not only in medicine, but also in many aspects of society that increasingly depend on data capture and processing. To make better decisions today, we must learn from the past, and the digital past is data. But integrating data has a higher cost, limiting this virtual perspective of the past. In this context, LLMs help integrate large-scale text data, while KGs help integrate large-scale structured data. Both approaches are complementary, and their combination can have many applications in different areas where data-driven decisions are made.”

Aidan Hogan

Clinical data interoperability

Clinical data interoperability is a major issue worldwide, and Chile is no exception. There have recently been legislative advances on this matter, aimed at promoting the portability of patient data and clinical records between different health institutions: the idea is that when a patient visits any clinic, hospital, or health center, it should be possible to share -respecting applicable privacy, use, and other rules and ethics-, their medical history in digital format, faithfully and completely, without mistakes or omissions that could even be fatal. “Currently, LLMs have a great capacity to integrate text from different sources, but they lack the precision needed for this use case. So we propose that LLMs need to be combined with other methods -particularly KGs-, so that they can be used in the context of clinical data interoperability, and other medical applications,” Hogan points out.

Source: https://digitalhealth.tu-dresden.de/llms-in-medicine-researchers-publish-solutions-to-increase-reproducibility-and-precision/