well-documented Alzheimer's dataset

Dataset Overview:

• Total MRI Images: The dataset includes scans from 457 individuals, each with 3 MRI scan NIfTI files.
• Format: MRI scans were extracted from NIfTI files, converted to PNG format, and processed for cleaner, more accurate analysis.
• Data Imbalance: The dataset contains an imbalance, so upsampling may be necessary based on specific research needs.
• Source: Sourced directly from the OASIS database, guaranteeing authenticity and easy citation.

Data Processing:

1. NIfTI to PNG Conversion: MRI scans from 457 individuals were extracted from 457*4 NIfTI files and converted into PNG format for easy integration into deep learning frameworks.
2. Skull Stripping: To enhance image quality, skull stripping was applied, removing non-brain tissues and focusing on brain anatomy.
3. Data Cleanup: I manually removed any images with black regions or incomplete brain displays, ensuring only high-quality, complete MRI scans remain.

Data Quality and Usability:

• Named Images: Each image is clearly named, allowing users to easily match it with the corresponding OASIS research phase and individual.
• Manual Cleanup: I carefully filtered out images with incomplete brain displays, ensuring only relevant data remains for research.
• Batch Processing Considerations: While converting to three-channel images for batch processing, be sure to delete any remaining files (such as ‘brain.nii’ or ‘mask.nii’) manually to avoid errors.

Applications:

• Alzheimer’s Diagnosis: This dataset is perfect for deep learning researchers aiming to improve the accuracy of Alzheimer’s diagnosis by training AI models on high-quality, well-labeled MRI scans.
• Medical Imaging: Ideal for developing medical imaging algorithms, especially those focused on detecting neurodegenerative diseases.
• Research Contribution: If well-received, I plan to release future datasets, including a skull-stripped dataset from ADNI, to further enhance Alzheimer’s research.