Datasets:

LuminaAI
/

Brain_Tumor_MRI_4c_Aug_Split

This dataset contains MRI images for classifying brain tumors across four categories: glioma, meningioma, pituitary tumor, and no tumor. Each image is stored as an individual .png file, and the dataset is structured to be compatible with Lumina AI's Random Contrast Learning (RCL) algorithm via the PrismRCL application.

Dataset Structure

The dataset is organized into the following structure:

brain-cancer-4class/
    train/
        glioma_tumor/
            image_001.png
            image_002.png
            ...
        meningioma_tumor/
            image_001.png
            image_002.png
            ...
        pituitary_tumor/
            image_001.png
            image_002.png
            ...
        no_tumor/
            image_001.png
            image_002.png
            ...
    test/
        glioma_tumor/
            image_001.png
            image_002.png
            ...
        meningioma_tumor/
            image_001.png
            image_002.png
            ...
        pituitary_tumor/
            image_001.png
            image_002.png
            ...
        no_tumor/
            image_001.png
            image_002.png
            ...

Note: All file names must be unique across class folders. Only .png format is supported by PrismRCL for image inputs.

Features

Image Data: Each file contains a brain MRI scan in .png format.
Classes: Four folders represent the classification targets:
- glioma_tumor
- meningioma_tumor
- pituitary_tumor
- no_tumor

Usage (pre-split; optimal parameters)

Here is an example of how to load the dataset using PrismRCL:

C:\PrismRCL\PrismRCL.exe chisquared rclticks=10 boxdown=0 ^
data=C:\path\to\brain-cancer-4class\train testdata=C:\path\to\brain-cancer-4class\test ^
savemodel=C:\path\to\models\brain_cancer_4class.classify ^
log=C:\path\to\log_files stopwhendone

Explanation of Command:

C:\PrismRCL\PrismRCL.exe: Path to the PrismRCL executable for classification
chisquared: Specifies Chi-squared as the training evaluation method
rclticks=10: Number of RCL iterations during training
boxdown=0: RCL training parameter
data=C:\path\to\brain-cancer-4class\train: Path to the training dataset
testdata=C:\path\to\brain-cancer-4class\test: Path to the testing dataset
savemodel=C:\path\to\models\brain_cancer_4class.classify: Output path for the trained model
log=C:\path\to\log_files: Log file storage directory
stopwhendone: Ends the session when training is complete

License

This dataset is licensed under the MIT License.

Original Source

This dataset was prepared from publicly available medical imaging archives and academic datasets. Please cite the appropriate dataset creators if you use this in research or applications. For example, similar datasets appear in:
Cheng, Jun, et al. "Enhanced performance of brain tumor classification via tumor region augmentation and partition." PloS One, 10.10 (2015): e0140381.

Additional Information

All files are preprocessed .png images to ensure compatibility with PrismRCL. No resizing or normalization is required when using PrismRCL version 2.4.0 or later.