How jan-dyndor/alzheimer-mri-detection works
This repository is a notebook-based research project for classifying Alzheimer’s disease from brain MRI scans. Its documented workflow prepares MRI data, compares patient-level and leakage-prone splits, trains CNNs across augmentation and classification variants, and evaluates their results. The README describes these stages and experiments, but the notebook source was not available for inspection; links therefore reflect documented data flow rather than verified function calls.
Main entry points: alzheimer-detection-using-mri-scans.ipynb Model training
Components 15
Data preparation
- MRI dataset
- Image preprocessing
notebooks/alzheimer-detection-using-mri-scans.ipynb - Patient-level split
notebooks/alzheimer-detection-using-mri-scans.ipynb - Image augmentation
notebooks/alzheimer-detection-using-mri-scans.ipynb
Model experiments
- Four-class task
notebooks/alzheimer-detection-using-mri-scans.ipynb - Binary task
notebooks/alzheimer-detection-using-mri-scans.ipynb - ResNet50
notebooks/alzheimer-detection-using-mri-scans.ipynb - VGG16
notebooks/alzheimer-detection-using-mri-scans.ipynb - EfficientNetB4
notebooks/alzheimer-detection-using-mri-scans.ipynb - Model training
notebooks/alzheimer-detection-using-mri-scans.ipynb - Leakage experiment
notebooks/alzheimer-detection-using-mri-scans.ipynb
Evaluation
- Model evaluation
notebooks/alzheimer-detection-using-mri-scans.ipynb - Results comparison
notebooks/alzheimer-detection-using-mri-scans.ipynb
Other components
- Researcher
- Research notebook
notebooks/alzheimer-detection-using-mri-scans.ipynb
Connections 16
- Researcher → Research notebook: runs
- MRI dataset → Image preprocessing: MRI images
- Image preprocessing → Patient-level split: prepared images
- Patient-level split → Image augmentation: training images
- Patient-level split → Four-class task: four labels
- Patient-level split → Binary task: binary labels
- Four-class task → Model training: task setup
- Binary task → Model training: task setup
- Image augmentation → Model training: augmented inputs
- ResNet50 → Model training: backbone
- VGG16 → Model training: backbone
- EfficientNetB4 → Model training: backbone
- Patient-level split → Leakage experiment: non-patient split
- Model training → Model evaluation: predictions
- Leakage experiment → Model evaluation: predictions
- Model evaluation → Results comparison: metrics