ROCO-Radiology: Medical Imaging Dataset with Associated Text

The ROCO-radiology dataset is a subset of the large-scale Radiology Objects in COntext (ROCO) collection, focusing on medical imaging. It was modified by the author mdwiratathya to select only radiology data and convert images to PIL objects. The dataset was last updated on the Hugging Face platform on June 14, 2024.

Use Cases

• Train image captioning models based on radiology images and associated text descriptions.
• Develop medical visual question answering (VQA) systems based on the multimodal context.
• Benchmark multimodal classification algorithms for radiology findings.
• Fine-tune vision-language models for domain-specific medical applications.

Strengths

• Derived from the large-scale ROCO dataset, which is a known medical multimodal collection.
• Focuses specifically on the radiology domain, a key medical imaging subfield.
• Images are provided as PIL objects, which may simplify initial loading for Python users.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license information are unknown, which may limit suitability assessment.

Provenance

Source

Derived from the Radiology Objects in COntext (ROCO) dataset.

Collection Method

Subset selected and images converted to PIL objects by the author.

Freshness

Last updated 2024-06-14 06:37:59; freshness should be verified.