Position: Transparent Reporting for Healthcare GenAI

Abstract

Generative AI systems in healthcare offer significant potential but face challenges due to the lack of standardized reporting frameworks. This gap hinders reproducibility, evaluation and regulation, potentially compromising patient safety. Existing tools like Data Cards and Model Cards are inadequate for healthcare’s specific needs, such as regulatory compliance and clinical relevance. Drawing on the success of the STROBE checklist in epidemiology, we propose an extension tailored for GenAI in healthcare. This framework enhances transparency, reproducibility and safety, supporting the effective integration of GenAI systems into clinical practice.

PDF NeurIPSW OpenReview Semantic Scholar

Cite

Text

Kolbeinsson and Kolbeinsson. "Position: Transparent Reporting for Healthcare GenAI." NeurIPS 2024 Workshops: GenAI4Health, 2024.

Markdown

[Kolbeinsson and Kolbeinsson. "Position: Transparent Reporting for Healthcare GenAI." NeurIPS 2024 Workshops: GenAI4Health, 2024.](https://mlanthology.org/neuripsw/2024/kolbeinsson2024neuripsw-position-a/)

BibTeX

@inproceedings{kolbeinsson2024neuripsw-position-a,
  title     = {{Position: Transparent Reporting for Healthcare GenAI}},
  author    = {Kolbeinsson, Arinbjörn and Kolbeinsson, Benedikt},
  booktitle = {NeurIPS 2024 Workshops: GenAI4Health},
  year      = {2024},
  url       = {https://mlanthology.org/neuripsw/2024/kolbeinsson2024neuripsw-position-a/}
}