Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • harvard-cite-them-right
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Benchmarking Process Model Representations for the LLM-Driven HealthProcessAI Framework
Karolinska Institute, Stockholm, Sweden.
Karolinska Institute, Stockholm, Sweden; Royal Institute of Technology; Sweden Karolinska University Hospital, Sweden.
University of Borås, Faculty of Textiles, Engineering and Business. University of Borås, Faculty of Caring Science, Work Life and Social Welfare. Karolinska Institute, Stockholm, Sweden; Karolinska University Hospital, Stockholm, Sweden.ORCID iD: 0000-0002-6995-967X
2026 (English)In: Move Beyond: 2026 11th International Conference on Smart and Sustainable Technologies (SpliTech), 23-26 June 2026, 2026, p. 1-6Conference paper, Published paper (Refereed)
Abstract [en]

In recent years, process mining has emerged as a powerful analytical technique for understanding and visualizing complex healthcare workflows. However, its application in real clinical environments faces technical complexity and limited access to practical training resources, making it difficult for clinicians to adopt this methodology. For this reason, we introduced HealthProcessAI, a GenAI framework designed to simplify process mining applications in healthcare for automated process map interpretation and report generation, helping clinicians understand process mining. However, there are still critical steps to take before moving on to clinical validation. This paper moves forward with the implementation of HealthProcessAI by performing a sensitivity analysis to understand the effect of the data format used to represent the process map in the generated report. The results show that Qwen3-vl:30b and Gemma3:27b - both with a score equal to 3.1/4.0 - emerged as the strongest performers in a local, privacy-preserving deployment setting, as well as the semantic enriched formats. Furthermore, images showed a high score with Qwen3-vl:30b and Gemma3:27b allowing to use our framework when text based formats are not available. The next steps should focus on the implementation of image to text conversors to improve their processing and clinical adoption

Place, publisher, year, edition, pages
2026. p. 1-6
Keywords [en]
Modeling, Process mining, Medical services, Resource description framework, Large language models, Printing, Process modeling, Artificial intelligence, Timing, Sensitivity analysis
National Category
Computer Sciences
Identifiers
URN: urn:nbn:se:hb:diva-36427DOI: 10.23919/splitech70133.2026.11685502Scopus ID: 2-s2.0-105052163796ISBN: 978-953-290-147-4 (electronic)OAI: oai:DiVA.org:hb-36427DiVA, id: diva2:2111571
Conference
2026 11th International Conference on Smart and Sustainable Technologies (SpliTech), 23-26 June 2026, Split, Croatia
Available from: 2026-10-20 Created: 2026-10-20 Last updated: 2026-10-20Bibliographically approved

Open Access in DiVA

No full text in DiVA

Other links

Publisher's full textScopus

Authority records

Seoane, Fernando

Search in DiVA

By author/editor
Seoane, Fernando
By organisation
Faculty of Textiles, Engineering and BusinessFaculty of Caring Science, Work Life and Social Welfare
Computer Sciences

Search outside of DiVA

GoogleGoogle Scholar

doi
isbn
urn-nbn

Altmetric score

doi
isbn
urn-nbn
Total: 15 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • harvard-cite-them-right
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf