Pathology Image Question-Answer Pairs from Medical Textbooks

PathVQA is a dataset for Medical Visual Question Answering built from the 'Textbook of Pathology' and 'Basic Pathology' textbooks. It contains question-answer pairs on pathology images, including both open-ended and binary yes/no questions.

Use Cases

• Train a VQA model on pathology images with open-ended and binary question-answer pairs.
• Benchmark medical AI systems using question-answer pairs derived from pathology textbooks.
• Analyze the performance of vision-language models on specialized medical yes/no questions.

Strengths

• Specifically designed for Medical Visual Question Answering, a specialized AI task.
• Derived from two authoritative pathology textbooks: 'Textbook of Pathology' and 'Basic Pathology'.
• Includes a mix of open-ended and binary yes/no question formats.

Limitations

• Dataset size, row count, and specific column structure are unknown.
• Limited to pathology images and text from two specific textbook sources, which may not represent all medical domains.
• Potential for label noise or bias inherent in the source textbook materials.

Provenance

Source

Built from 'Textbook of Pathology', 'Basic Pathology', and a public digital pathology library.

Collection Method

Question-answer pairs generated from pathology textbooks and images.

Freshness

Last updated on 2023-06-03.