BJyotibrat/Masoud-Nickparvar-Brain-Tumor-MRI-Dataset

Dataset Card for Brain Tumor MRI Dataset

<!-- Provide a quick summary of the dataset. -->

A dataset of 7,200 human brain MRI images, labeled into four classes — glioma, meningioma, pituitary tumor, and no tumor — for training and evaluating brain tumor classification models. This is a direct upload of the Brain Tumor MRI Dataset originally published on Kaggle by Masoud Nickparvar.

Dataset Details

Dataset Description

<!-- Provide a longer summary of what this dataset is. -->

This dataset combines MRI images from three source datasets — figshare, the SARTAJ dataset, and Br35H — into a single collection for multi-class brain tumor classification. It contains 7,200 T1-weighted contrast-enhanced brain MRI images, organized into four classes: glioma, meningioma, pituitary tumor, and no tumor.

The "no tumor" class images are sourced entirely from the Br35H dataset. Glioma class images were taken from the figshare dataset rather than SARTAJ, since the original dataset creator identified mislabeling issues in SARTAJ's glioma images and replaced them accordingly.

The dataset is pre-split into Training (5,600 images) and Testing (1,600 images) directories, with each further divided into the four class subfolders. Image sizes vary across the dataset, so resizing/preprocessing is recommended before model training.

• —Curated by: Masoud Nickparvar (original Kaggle dataset creator); combined from figshare, SARTAJ, and Br35H source datasets
• —Shared by: BJyotibrat
• —Language(s) (NLP): Not applicable (image dataset; class labels in English)
• —License: CC BY 4.0

Dataset Sources

<!-- Provide the basic links for the dataset. -->

• —Original Kaggle Dataset: Brain Tumor MRI Dataset
• —Source Datasets: figshare brain tumor dataset, SARTAJ brain tumor classification dataset, Br35H brain tumor detection dataset

Uses

<!-- Address questions around how the dataset is intended to be used. -->

Direct Use

<!-- This section describes suitable use cases for the dataset. -->

Intended for training, fine-tuning, and evaluating image classification models for brain tumor detection and classification research and education, including transfer learning experiments on architectures such as EfficientNet, ResNet, and other CNN-based or vision transformer models.

Out-of-Scope Use

<!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->

This dataset must not be used to train or deploy models intended for real clinical diagnosis without proper medical validation, regulatory approval, and oversight by licensed medical professionals. It is not suitable for tasks requiring precise tumor localization or segmentation (it provides classification labels only, not pixel-level annotations), and it should not be treated as demographically or clinically representative of any specific patient population, since detailed patient metadata is not provided.

Dataset Structure

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->

The dataset is organized as image files under two top-level splits, each containing four class subfolders:

Training/
├── glioma/
├── meningioma/
├── notumor/
└── pituitary/
Testing/
├── glioma/
├── meningioma/
├── notumor/
└── pituitary/

Approximate class distribution across the full dataset:

Split sizes: 5,600 images in Training, 1,600 images in Testing (roughly an 80/20 split). Image dimensions vary across samples; no fixed resolution is enforced in the raw files.

Dataset Creation

Curation Rationale

<!-- Motivation for the creation of this dataset. -->

The dataset was curated to provide a larger, more diverse, and more reliably labeled collection for brain tumor classification research than any single source dataset alone. The original creator specifically addressed a known labeling issue in the SARTAJ dataset's glioma class by substituting those images with correctly labeled glioma images from figshare.

Source Data

<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->

Data Collection and Processing

<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->

The dataset combines three previously published, independently collected MRI datasets:

• —figshare — 3,064 T1-weighted contrast-enhanced MRI slices from 233 patients, covering meningioma, glioma, and pituitary tumor classes.
• —SARTAJ dataset — used as a source for meningioma, pituitary, and no-tumor images (glioma images from this source were excluded due to labeling issues).
• —Br35H — used as the source for all "no tumor" class images.

No additional preprocessing (resizing, normalization, augmentation) has been applied to the raw images in this dataset; users are expected to preprocess images as needed for their specific modeling pipeline.

Who are the source data producers?

<!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->

The original MRI scans were produced by the creators of the figshare, SARTAJ, and Br35H datasets respectively. No further demographic information about patients or imaging institutions is provided by the original dataset sources.

Annotations

<!-- If the dataset contains annotations which are not part of the initial data collection, use this section to describe them. -->

Annotation process

<!-- This section describes the annotation process such as annotation tools used in the process, the amount of data annotated, annotation guidelines provided to the annotators, interannotator statistics, annotation validation, etc. -->

Class labels (glioma, meningioma, pituitary, no tumor) were inherited from the three source datasets' original annotations. The dataset curator performed additional validation, identifying and correcting a labeling inconsistency in the SARTAJ dataset's glioma class by replacing those images with figshare's glioma images.

Who are the annotators?

<!-- This section describes the people or systems who created the annotations. -->

Original annotations were produced by the creators/maintainers of the figshare, SARTAJ, and Br35H datasets.

Personal and Sensitive Information

<!-- State whether the dataset contains data that might be considered personal, sensitive, or private (e.g., data that reveals addresses, uniquely identifiable names or aliases, racial or ethnic origins, sexual orientations, religious beliefs, political opinions, financial or health data, etc.). If efforts were made to anonymize the data, describe the anonymization process. -->

This dataset consists of medical brain MRI scans, which are inherently sensitive health data. Images are anonymized in the sense that no patient names, dates of birth, or other direct identifiers are attached to the files or folder structure. However, MRI scans can in principle be used for re-identification in combination with other data sources; no explicit de-identification methodology beyond removal of file-level metadata is documented by the original source datasets.

Bias, Risks, and Limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->

• —No patient demographic metadata: The dataset does not include age, sex, ethnicity, scanner type, or acquisition site information, making it impossible to assess or correct for demographic representation gaps.
• —Combined-source inconsistency: Since images come from three separate datasets with potentially different scanners, protocols, and imaging conditions, models trained on this data may pick up on source-specific artifacts rather than purely tumor-relevant features.
• —Label reliability: While the glioma labeling issue from SARTAJ was addressed by the curator, other potential labeling inaccuracies inherited from the source datasets have not been independently re-verified.
• —Not a diagnostic-grade dataset: This dataset is intended for research and educational purposes. It has not been validated for clinical or diagnostic use, and models trained on it should not be deployed for real patient care without further validation and regulatory approval.
• —Classification only, no segmentation: The dataset provides image-level class labels, not tumor location, size, or pixel-level segmentation masks.

Recommendations

<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->

Users should be made aware of the risks, biases, and limitations of the dataset. In particular:

• —Do not use models trained on this dataset for real clinical diagnosis without proper validation, regulatory approval, and oversight by licensed medical professionals.
• —Be aware of potential source-dataset bias (scanner/protocol differences across figshare, SARTAJ, and Br35H).
• —Treat this as a benchmark/research dataset rather than a clinically representative sample.

Citation

<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->

BibTeX:

@misc{msoud_nickparvar_2026,
	title={Brain Tumor MRI Dataset},
	url={https://www.kaggle.com/dsv/14832123},
	DOI={10.34740/KAGGLE/DSV/14832123},
	publisher={Kaggle},
	author={Msoud Nickparvar},
	year={2026}
}

APA:

Msoud Nickparvar. (2026). Brain Tumor MRI Dataset [Dataset]. Kaggle. https://doi.org/10.34740/KAGGLE/DSV/14832123

More Information

This is a direct re-upload of the original Kaggle dataset, made available on Hugging Face for convenience of access. This dataset is licensed under CC BY 4.0, which requires attribution to the original creator when the dataset (or derivative works from it) is used or redistributed — see the Citation section above for the required attribution format. All credit for dataset curation belongs to the original creator, Masoud Nickparvar, and the maintainers of the figshare, SARTAJ, and Br35H source datasets. This dataset card's uploader has not modified the underlying image data.

Dataset Card Authors

Bindupautra Jyotibrat (BJyotibrat)

Dataset Card Contact

Email: bjyotibrat@gmail.com