What type of study is this?

This is a Quantitative Study study.

October 3, 2025Open Access

MOSAIC: A Multilingual, Taxonomy-Agnostic, and Computationally Efficient Approach for Radiological Report Classification

View Full Paper

Ask AI

Bookmark

Authors

MOSAIC demonstrates high performance in radiological report classification across multiple languages, highlighting its versatility.

MOSAIC achieves a mean macro F1 score of 88 across five chest X-ray datasets, indicating expert-level performance.
Using as few as 80 annotated samples, MOSAIC can still reach a weighted F1 score of 82 on Danish reports, demonstrating efficiency.
The model is built on MedGemma-4B, a compact open-access language model, allowing easy deployment on consumer-grade GPUs.
MOSAIC provides an alternative to large proprietary models, encouraging contribution from the community for further enhancements.

Schiavone et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark