Julian Varghese

Prof. Dr. Julian Varghese

Institute of Medical Informatics, University of Münster, 48149 Münster, Germany
orcid

Research Focus & Metrics

Academic Career Length

14Years (2013-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
23 April 2025 in Nature Medicine
Verified
Journal Article
Open access
PDF file
SSSarah Sandmann
Sarah Sandmann
SHStefan Hegselmann
Stefan Hegselmann
MFMichael Fujarski
Michael Fujarski
LBLucas Bickmann
Lucas Bickmann
BWBenjamin Wild
Benjamin Wild
Julian VargheseJulian Varghese
Julian Varghese
Large Language Models (LLMs) are increasingly transforming medical applications. However, proprietary models such as GPT-4o face significant barriers to clinical adoption because they cannot be deployed on site within healthcare institutions, making them non-compliant with stringent privacy regulations. Recent advancements in open-source LLMs such as DeepSeek models offer a promising alternative since they allow efficient fine-tuning on local data in hospitals with advanced IT infrastructure. To demonstrate the clinical utility of DeepSeek-V3 and DeepSeek-R1, we benchmarked their performance on clinical decision support tasks against proprietary LLMs, including GPT-4o and Gemini-2.0 Flash Thinking Experimental. Using 125 patient cases with sufficient statistical power, covering a broad range of frequent and rare diseases, we found that DeepSeek models perform equally well and in some cases better than proprietary LLMs. Our study demonstrates that open-source LLMs can provide a scalable pathway for secure model training enabling real-world medical applications in accordance with data privacy and healthcare regulations.

6 March 2024 in Nature Communications
Verified
Journal Article
Open access
PDF file
SSSarah Sandmann
Sarah Sandmann
SRSarah Riepenhausen
Sarah Riepenhausen
LPLucas Plagwitz
Lucas Plagwitz
Julian VargheseJulian Varghese
Julian Varghese
It is likely that individuals are turning to Large Language Models (LLMs) to seek health advice, much like searching for diagnoses on Google. We evaluate clinical accuracy of GPT-3·5 and GPT-4 for suggesting initial diagnosis, examination steps and treatment of 110 medical cases across diverse clinical disciplines. Moreover, two model configurations of the Llama 2 open source LLMs are assessed in a sub-study. For benchmarking the diagnostic task, we conduct a naïve Google search for comparison. Overall, GPT-4 performed best with superior performances over GPT-3·5 considering diagnosis and examination and superior performance over Google for diagnosis. Except for treatment, better performance on frequent vs rare diseases is evident for all three approaches. The sub-study indicates slightly lower performances for Llama models. In conclusion, the commercial LLMs show growing potential for medical question answering in two successive major releases. However, some weaknesses underscore the need for robust and regulated AI models in health care. Open source LLMs can be a viable option to address specific needs regarding data privacy and transparency of training.

23 September 2017 in Journal of the American Medical Informatics Association
Verified
Journal Article
Closed access
PDF file
Julian VargheseJulian Varghese
Julian Varghese
MKMaren Kleine
Maren Kleine
SGSophia Isabella Gessner
Sophia Isabella Gessner
SSSarah Sandmann
Sarah Sandmann
MDMartin Dugas
Martin Dugas
To systematically classify the clinical impact of computerized clinical decision support systems (CDSSs) in inpatient care. Medline, Cochrane Trials, and Cochrane Reviews were searched for CDSS studies that assessed patient outcomes in inpatient settings. For each study, 2 physicians independently mapped patient outcome effects to a predefined medical effect score to assess the clinical impact of reported outcome effects. Disagreements were measured by using weighted kappa and solved by consensus. An example set of promising disease entities was generated based on medical effect scores and risk of bias assessment. To summarize technical characteristics of the systems, reported input variables and algorithm types were extracted as well. Seventy studies were included. Five (7%) reported reduced mortality, 16 (23%) reduced life-threatening events, and 28 (40%) reduced non–life-threatening events, 20 (29%) had no significant impact on patient outcomes, and 1 showed a negative effect (weighted κ: 0.72, P < .001). Six of 24 disease entity settings showed high effect scores with medium or low risk of bias: blood glucose management, blood transfusion management, physiologic deterioration prevention, pressure ulcer prevention, acute kidney injury prevention, and venous thromboembolism prophylaxis. Most of the implemented algorithms (72%) were rule-based. Reported input variables are shared as standardized models on a metadata repository. Most of the included CDSS studies were associated with positive patient outcomes effects but with substantial differences regarding the clinical impact. A subset of 6 disease entities could be filtered in which CDSS should be given special consideration at sites where computer-assisted decision-making is deemed to be underutilized. on PROSPERO: CRD42016049946.

12 October 2020 in Visceral Medicine
Verified
Journal Article
Open access
PDF file
Julian VargheseJulian Varghese
Julian Varghese
Background: Artificial intelligence (AI) applications that utilize machine learning are on the rise in clinical research and provide highly promising applications in specific use cases. However, wide clinical adoption remains far off. This review reflects on common barriers and current solution approaches. Summary: Key challenges are abbreviated as the RISE criteria: Regulatory aspects, Interpretability, interoperability, and the need for Structured data and Evidence. As reoccurring barriers of AI adoption, these concepts are delineated and complemented by points to consider and possible solutions for effective and safe use of AI applications. Key Messages: There is a fraction of AI applications with proven clinical benefits and regulatory approval. Many new promising systems are the subject of current research but share common issues for wide clinical adoption. The RISE criteria can support preparation for challenges and pitfalls when designing or introducing AI applications into clinical practice.

11 January 2022 in Journal of Medical Internet Research
Verified
Journal Article
Open access
HUHannes Ulrich
Hannes Ulrich
AKAnn-Kristin Kock-Schoppenhauer
Ann-Kristin Kock-Schoppenhauer
NDNoemi Deppenwiese
Noemi Deppenwiese
RGRobert Gött
Robert Gött
JKJori Kern
Jori Kern
JIJosef Ingenerf
Josef Ingenerf
Background: Metadata are created to describe the corresponding data in a detailed and unambiguous way and is used for various applications in different research areas, for example, data identification and classification. However, a clear definition of metadata is crucial for further use. Unfortunately, extensive experience with the processing and management of metadata has shown that the term “metadata” and its use is not always unambiguous. Objective: This study aimed to understand the definition of metadata and the challenges resulting from metadata reuse. Methods: A systematic literature search was performed in this study following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines for reporting on systematic reviews. Five research questions were identified to streamline the review process, addressing metadata characteristics, metadata standards, use cases, and problems encountered. This review was preceded by a harmonization process to achieve a general understanding of the terms used. Results: The harmonization process resulted in a clear set of definitions for metadata processing focusing on data integration. The following literature review was conducted by 10 reviewers with different backgrounds and using the harmonized definitions. This study included 81 peer-reviewed papers from the last decade after applying various filtering steps to identify the most relevant papers. The 5 research questions could be answered, resulting in a broad overview of the standards, use cases, problems, and corresponding solutions for the application of metadata in different research areas. Conclusions: Metadata can be a powerful tool for identifying, describing, and processing information, but its meaningful creation is costly and challenging. This review process uncovered many standards, use cases, problems, and solutions for dealing with metadata. The presented harmonized definitions and the new schema have the potential to improve the classification and generation of metadata by creating a shared understanding of metadata and its context.

6 August 2023 in Journal of Hepatology
Verified
Journal Article
Open access
Julian VargheseJulian Varghese
Julian Varghese
JCJulius Chapiro
Julius Chapiro
In an age where technology is evolving at a sometimes incomprehensibly rapid pace, the liver community must adjust and learn to embrace breakthroughs with an open mind in order to benefit from potentially transformative influences on our science and practice. The Journal of Hepatology has responded to novel developments in artificial intelligence (AI) by recruiting experts in the field to serve on the Editorial Board. Publications introducing novel AI technology are no longer uncommon in our journal and are among the most highly debated and possibly practice-changing papers across a broad range of scientific disciplines, united by their focus on liver disease. As AI is rapidly evolving, this expert paper will focus on educating our readership on large language models and their possible impact on our research practice and clinical outlook, outlining both challenges and opportunities in the field. "To improve is to change; to be perfect is to change often." - Winston S. Churchill.