Automating Performance Status Annotation in Oncology Using Llama-3.

Journal: Studies in health technology and informatics
Published Date:

Abstract

This work explores the automated extraction of medical information from Dutch clinical notes using Llama-3 and a limited amount of annotations. We compared zero-, one- and few-shot learning for the extraction of performance status of patients with palliative esophagogastric cancer few-shot learning and one-shot with ACSESS-selected examples, showed the best performance. Future work shall focus on improving the model's precision.

Authors

  • Irene Cara
    Netherlands Comprehensive Cancer Organisation (IKNL), Utrecht.
  • Nynke van 't Hof
    Netherlands Comprehensive Cancer Organisation (IKNL), Utrecht.
  • Sebastiaan Siegerink
    Amsterdam UMC, Department of Medical Oncology, Amsterdam.
  • Linde Veen
    Amsterdam UMC, Department of Medical Oncology, Amsterdam.
  • Gijs Geleijnse
    Netherlands Comprehensive Cancer Organization (IKNL), Eindhoven, NL.
  • Hanneke van Laarhoven
    Amsterdam UMC, Department of Medical Oncology, Amsterdam.
  • Rob Verhoeven
    Netherlands Comprehensive Cancer Organisation (IKNL), Utrecht.