Large language model-based uncertainty-adjusted label extraction for artificial intelligence model development in upper extremity radiography - Takeaways - MDSpire

Large language model-based uncertainty-adjusted label extraction for artificial intelligence model development in upper extremity radiography

  • By

  • Hanna Kreutzer

  • Anne-Sophie Caselitz

  • Thomas Dratsch

  • Daniel Pinto dos Santos

  • Christiane Kuhl

  • Daniel Truhn

  • Sven Nebelung

  • November 14, 2025

  • 0 min

Share

  • 1

    Automated label extraction using large language models (LLMs) can enhance data availability for AI model training in upper extremity radiography.

  • 2

    Traditional manual annotation methods for radiologic reports are labor-intensive and inconsistent, leading to potential mislabeling in datasets.

  • 3

    This study investigates the effectiveness of LLMs in extracting labels while accounting for diagnostic uncertainty in radiology reports.

  • 4

    The research utilizes a two-center retrospective analysis to validate the performance of LLMs in multi-label classification models.

  • 5

    Findings suggest that LLMs can accurately extract labels and that label uncertainty may not significantly impact model performance.

Original Source(s)

Large language model-based uncertainty-adjusted label extraction for artificial intelligence model development in upper extremity radiography

  • by Hanna Kreutzer, Anne-Sophie Caselitz, Thomas Dratsch, Daniel Pinto dos Santos, Christiane Kuhl, Daniel Truhn, Sven Nebelung

  • November 14, 2025

Related Content

Diagnostic accuracy of deep learning vs. human raters for detecting osteoporotic vertebral compression fractures in routine CT scans

UroFusion-X: a unified multimodal deep learning framework for robust diagnosis, subtyping, and prognosis of urological cancers

Trauma CT and Unexpected Findings

Investigators evaluate classification of incidental findings on trauma whole-body CT.