ColonINST-v1: A Large-Scale Multimodal Instruction Dataset for Colonoscopy

A large-scale multimodal instruction tuning dataset for colonoscopy research, comprising over 300,000 colonoscopic images and 128,000 medical captions generated by GPT-4V. The dataset includes 62 categories and is designed to instruct models to execute user-driven tasks interactively. It was created by ai4colonoscopy and last updated on February 4, 2025.

Use Cases

• Training multimodal AI models for medical image captioning based on GPT-4V generated descriptions.
• Developing interactive diagnostic assistants based on the dataset's instruction-response structure.
• Fine-tuning vision-language models for polyp detection and classification based on the 62 categories of colonoscopic images.
• Researching human-AI interaction in clinical settings based on the user-driven task execution framework.

Strengths

• Contains over 300,000 colonoscopic images, indicating a large-scale collection.
• Includes 128,000 medical captions generated by GPT-4V, providing detailed textual annotations.
• Covers 62 distinct categories, suggesting a structured and varied classification of medical conditions or findings.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific model training needs.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

ai4colonoscopy

Collection Method

Likely collected and annotated from medical sources, with captions generated by GPT-4V.

Freshness

Last updated 2025-02-04 17:22:01; freshness should be verified.

Geography

Regionus platform tag suggests a likely focus on the United States.