Medical-Vision: High-Quality Medical Visual Question Answering Dataset (Aquiles-ai/Medical-Vision)

Dataset Description

Medical-Vision is a curated dataset designed for Visual Question Answering (VQA) in medical contexts. This dataset combines high-quality medical images with corresponding questions and expert answers, making it ideal for training and evaluating vision-language models in healthcare applications.

Key Features

• 8,035 high-quality examples
• Diverse medical imaging modalities (X-rays, CT scans, MRI, pathology slides, etc.)
• Natural question-answer pairs covering clinical interpretations, diagnoses, and medical descriptions
• Carefully curated and preprocessed from multiple authoritative sources
• Randomly shuffled to prevent training biases

Dataset Structure

Data Fields

• image: PIL Image object containing the medical image
• question: String with the medical question about the image
• answer: String with the expert answer or description

Data Splits

Split	Examples
train	8,035

Usage

Loading the Dataset

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("Aquiles-ai/Medical-Vision")

# Access the training split
train_data = dataset['train']

# View dataset info
print(f"Number of examples: {len(train_data)}")
print(f"Features: {train_data.features}")

Example Usage

from datasets import load_dataset
from PIL import Image
import matplotlib.pyplot as plt

# Load dataset
dataset = load_dataset("Aquiles-ai/Medical-Vision")

# Get a random example
example = dataset['train'][0]

# Display the image
plt.figure(figsize=(10, 6))
plt.imshow(example['image'])
plt.axis('off')
plt.title('Medical Image')
plt.show()

# Print Q&A
print(f"Question: {example['question']}")
print(f"\nAnswer: {example['answer']}")

Output Example:

Question: What do you see in this image? Describe it medically.

Answer: The chest X-ray shows bilateral infiltrates consistent with 
pulmonary edema. There is also cardiomegaly with an enlarged cardiac 
silhouette. The costophrenic angles are preserved, and no pleural 
effusion is visible.

Applications

This dataset is suitable for:

• Medical Visual Question Answering: Training models to answer questions about medical images
• Clinical Decision Support: Developing AI assistants for radiologists and clinicians
• Medical Education: Creating interactive learning tools for medical students
• Vision-Language Models: Fine-tuning multimodal models (LLaVA, Qwen-VL, Asclepio, etc.)
• Medical Image Captioning: Generating descriptive captions for medical images

Dataset Creation

Quality Assurance

• Manual verification of image-question-answer alignment
• Removal of duplicates and low-quality examples
• Validation of image loading and accessibility
• Consistency checks across all data fields

Considerations for Use

Intended Use

This dataset is intended for:

• Research in medical AI and computer vision
• Development of clinical decision support tools
• Educational purposes in medical AI
• Fine-tuning vision-language models for healthcare

Limitations

• Not for clinical diagnosis: This dataset is for research and development only
• Language: Currently only available in English
• Image quality: Varies across source datasets
• Medical scope: May not cover all medical specialties equally
• Requires expert validation: Any clinical application requires validation by medical professionals

Ethical Considerations

• All images are from publicly available medical datasets
• No patient identifiable information (PII) is included
• Users should follow appropriate ethical guidelines when deploying models trained on this data
• Medical AI outputs should always be reviewed by qualified healthcare professionals

Citation

If you use this dataset in your research, please cite:

@dataset{medical_vision_2025,
  title={Medical-Vision: High-Quality Medical Visual Question Answering Dataset (Aquiles-ai/Medical-Vision)},
  author={Aquiles-ai},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/Aquiles-ai/Medical-Vision}
}

License

This dataset is released under the Apache 2.0 License. Please refer to individual source datasets for their specific licensing terms.

Contact

For questions, issues, or contributions, please open an issue on the dataset repository or contact the maintainers.

• More about Aquiles-ai.

• Aquiles-ai on GitHub.

• Our collections at HuggingFace.

Disclaimer: This dataset is provided for research and educational purposes only. It should not be used as a substitute for professional medical advice, diagnosis, or treatment.