BrainTumorClassification

A large-scale curated brain MRI image dataset for multi-class brain tumor classification and medical imaging research.

Dataset Overview

BrainTumorClassification is a curated collection of 16,599 brain MRI images designed for training and evaluating machine learning and deep learning models for brain tumor classification.

The dataset contains four diagnostic categories:

• Glioma
• Meningioma
• Pituitary Tumor
• No Tumor

This dataset was developed through the curation, validation, and reorganization of publicly available Brain Tumor MRI datasets, primarily derived from widely used Kaggle sources. Additional quality checks, data verification, and structural improvements were performed to create a larger and more accessible resource for the research community.

Dataset Structure

BrainTumorClassification/
│
├── train/
│   ├── _classes.csv
│   ├── part1/
│   │   └── MRI images
│   │
│   └── part2/
│       └── MRI images
│
├── valid/
│   ├── _classes.csv
│   └── MRI images
│
└── test/
    ├── _classes.csv
    └── MRI images

Train Split

The training split contains the majority of images and is divided into part1 and part2 directories. Both folders contain MRI images used for model training, while _classes.csv stores the corresponding label information.

Validation Split

The validation split contains MRI images used for model selection, hyperparameter tuning, and performance monitoring during training. Labels are provided through the accompanying _classes.csv file.

Test Split

The test split contains MRI images reserved for final evaluation and benchmarking. Labels are provided through the corresponding _classes.csv file.

Dataset Statistics

Property	Value
Total Images	16,599
Classes	4
Task	Image Classification
Domain	Medical Imaging
Modality	Brain MRI
License	MIT

Classification Labels

Label	Description
Glioma	Brain MRI scans containing Glioma tumors
Meningioma	Brain MRI scans containing Meningioma tumors
Pituitary	Brain MRI scans containing Pituitary tumors
No Tumor	Brain MRI scans without detectable tumor presence

Data Curation

This dataset was created through a structured curation pipeline that included:

• Collection from publicly available Brain Tumor MRI datasets.
• Verification of image-label consistency.
• Removal of corrupted or unusable files.
• Reorganization into a standardized directory structure.
• Quality validation and dataset cleaning.
• Preparation for machine learning and deep learning workflows.

The goal of this effort was to provide a larger, cleaner, and easier-to-use dataset for researchers, students, and practitioners working on medical image classification.

Intended Uses

This dataset is suitable for:

• Brain tumor classification
• Deep learning research
• Computer vision applications
• Transfer learning experiments
• Vision Transformer (ViT) training
• Explainable AI (XAI)
• Medical imaging research
• Academic projects and benchmarking

Citation

If you use this dataset in your research, publication, thesis, report, or product, please cite:

@dataset{patel2025braintumorclassification,
  author = {Dev Patel},
  title = {BrainTumorClassification: A Curated Brain MRI Classification Dataset},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/devpatel1012/BrainTumorClassification}
}

---

Curated by Dev Patel

Total Images: 16,599

Task: Multi-Class Brain Tumor Classification