Musculoskeletal Radiology Reports and LLM Diagnostic Extraction Evaluation

583.0 KB of supplementary data for a 2026 study evaluating ChatGPT-4.0 for diagnostic extraction from musculoskeletal radiology reports. The data includes original reports, a survey for human readers, collected answers from humans and the AI, and Python evaluation code. It was authored by Wolfram A. Bosbach and published under a CC-BY-4.0 license.

Use Cases

• Benchmarking large language model performance on clinical information extraction based on the provided human-AI comparison data.
• Training or fine-tuning models for diagnostic entity recognition from radiology reports based on the annotated report data.
• Studying inter-rater agreement and consensus in medical report interpretation based on the survey and collected answers from multiple human readers.

Strengths

• Includes direct comparison data between human expert readers and a large language model (ChatGPT-4.0).
• Provides the original radiology reports and the Python code used for evaluation, supporting reproducibility.
• Released under a permissive CC-BY-4.0 license for open use.

Limitations

• The dataset is small (583.0 KB), indicating limited scope and sample size.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Supplementary material for a peer-reviewed publication in the journal 'Diagnostics'.

Collection Method

Likely collected as part of a research study involving human reader surveys and AI model outputs.

Freshness

Last updated 2026-05-22 17:58:52.