Ultrasound Radiology Reports For Language Model Fine-Tuning

A text dataset of radiology reports specifically for ultrasound examinations, intended for fine-tuning large language models. The dataset was created by Ka4on and uploaded to Hugging Face in October 2023. The exact size, scope, and number of reports are not specified.

Use Cases

• Fine-tune a language model to generate ultrasound report findings based on input exam notes.
• Train a model to classify report sections (e.g., indication, findings, impression) within ultrasound documents.
• Develop a named entity recognition system to extract key clinical entities like anatomical locations or measurements from ultrasound report text.

Strengths

• Dataset is specifically curated for the focused task of fine-tuning LLMs on radiology reports.
• Provides a direct application for AI in medical imaging text analysis.

Limitations

• No quantitative details (row count, report count, word count) are provided to assess scale.
• The source, collection methodology, and potential biases in the report data are unknown.
• Limited to the single modality of ultrasound, not covering other imaging types.

Provenance

Source

Hugging Face user Ka4on; original data source is unspecified.

Collection Method

Presumably collected or compiled from ultrasound radiology reports; exact method is null.

Freshness

Last updated on the Hugging Face platform in October 2023.