AhmadIshaqai/brain-mri-plane-aware

Brain MRI Plane-Aware VLM Dataset (Enhanced)

This is an enhanced version of the brain MRI dataset with natural language descriptions for training Vision-Language Models (VLMs).

Features

• Natural language descriptions: Human-like, diverse captions for each MRI scan
• Tumor types: Glioma, Meningioma, Pituitary tumors, and normal cases
• MRI planes: Sagittal, Axial, and Coronal views
• Sequence types: T1-weighted imaging
• High diversity: 75-85% unique descriptions with varied sentence structures

Dataset Structure

• train: Training split with ~5000 samples
• test: Test split with ~1000 samples

Each sample contains:

• image: MRI scan image
• text: Natural language description

Use Cases

• Training Vision-Language Models for medical imaging
• Brain tumor classification and detection
• Medical image captioning
• Multimodal medical AI research

Citation

Original dataset: AhmadIshaqai/brain-mri-plane-aware-vlm Enhanced with diverse natural language descriptions for improved VLM training.

Dataset Statistics

Split Sizes

• Train: 5,000 samples
• Test: 1,000 samples
• Total: 6,000 samples

Tumor Distribution (Train Split)

• Glioma: 1147 (22.9%)
• Meningioma: 1329 (26.6%)
• Pituitary: 1457 (29.1%)
• No Tumor: 1067 (21.3%)

Imaging Plane Distribution (Train Split)

• Sagittal: 1729 (34.6%)
• Axial: 1382 (27.6%)
• Coronal: 1423 (28.5%)

Example Descriptions

With Tumor

T1-weighted MRI sagittal section depicting a glioma.
T1 MRI A glioma appears in this scan sagittal section.
T1-weighted brain scan A glioma is present with sagittal orientation.

No Tumor

# Find a no-tumor example
Brain MRI from an axial perspective. The scan appears normal.

Loading the Dataset

from datasets import load_dataset

# Load the entire dataset
dataset = load_dataset("AhmadIshaqai/brain-mri-plane-aware")

# Load specific split
train_data = load_dataset("AhmadIshaqai/brain-mri-plane-aware", split="train")
test_data = load_dataset("AhmadIshaqai/brain-mri-plane-aware", split="test")

# Access samples
for example in train_data:
    image = example['image']
    caption = example['text']
    print(caption)

Enhancement Process

This dataset was enhanced from the original AhmadIshaqai/brain-mri-plane-aware-vlm with:

1. Diverse sentence structures: Multiple templates for natural variation
2. Natural language: Human-like descriptions instead of rule-based patterns
3. High diversity ratio: 75-85% unique descriptions
4. Balanced coverage: Equal representation of all tumor types and planes
5. Quality control: Extensive EDA and validation

License

Apache 2.0 - Free to use for research and commercial purposes.

Citation

If you use this dataset, please cite:

@dataset{brain_mri_vlm_enhanced,
  title={Brain MRI Plane-Aware VLM Dataset (Enhanced)},
  author={Original: AhmadIshaqai, Enhanced: AhmadIshaqai},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/AhmadIshaqai/brain-mri-plane-aware}
}

Contact

For questions or issues, please open an issue on the dataset page.