Augmented Alzheimer MRI 4-Classes Dataset
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
This dataset contains 12,800 preprocessed and augmented Magnetic Resonance Imaging (MRI) scans specifically designed for the classification of Alzheimer's Disease (AD) progression stages. The data is categorized into four distinct clinical stages: NonDemented (ND), VeryMildDemented (VMD), MildDemented (MD), and ModerateDemented (MOD). The raw MRI images were originally sourced from the Alzheimer's Disease Neuroimaging Initiative (ADNI) via the Kaggle platform. Because the original dataset suffers from severe class imbalance (heavily skewed towards the NonDemented class), this dataset provides a rigorously balanced alternative. A stratified data augmentation pipeline—incorporating geometric transformations (random rotations, translations) and photometric adjustments (brightness tuning)—was applied with intensity inversely proportional to each class's initial sample size. The resulting dataset achieves perfect class parity, containing exactly 3,200 images per class. This balanced distribution mitigates the risk of majority-class bias and is highly suitable for training, validating, and testing deep learning architectures (such as EfficientNet models) for early and accurate AD diagnosis.
Fields of Study & Subjects
Dementia and Cognitive Impairment Research
Retinal Imaging and Analysis
Medical Image Segmentation Techniques
Publication details
- Journal / source
- Mendeley Data
- Publication date
- 1 April 2026
- Language
- EN
- Document type
- Dataset