Alzheimer’s Disease Classification Using MRI Images Project Overview
This project focuses on classifying Alzheimer’s disease stages from brain MRI scan images using deep learning. The goal is to build a multi-class image classification model that can distinguish between different stages of Alzheimer’s disease to support early diagnosis and research.
The model classifies MRI images into four categories:
Non-Demented
Very Mild Demented
Mild Demented
Moderate Demented
Dataset Description
The dataset consists of MRI brain scan images organized into class-specific folders.
Classes & Image Count:
NonDemented: 12,800
VeryMildDemented: 11,200
MildDemented: 10,000
ModerateDemented: 10,000
Although the dataset is not perfectly balanced, class imbalance was handled during training using appropriate techniques (e.g., class weighting).
Source: Kaggle – Alzheimer’s Multiclass Dataset (Equal & Augmented)
Exploratory Data Analysis (EDA)
Basic EDA was performed to:
Inspect class distribution
Verify image quality and consistency
Confirm folder structure and labels
Model & Approach
Model Type: Convolutional Neural Network (CNN)
Technique: Transfer Learning
Base Model: Pretrained CNN (fine-tuned)
Loss Function: Categorical Crossentropy
Evaluation Metrics:
Accuracy
Precision
Recall
Model Performance The model was trained using transfer learning with fine-tuning and evaluated using multiple performance metrics.
Best Validation Results Validation Accuracy: 96.23% Validation AUC: ~0.995 Validation Precision: ~96.1% Validation Recall: ~95.8% Training Insights The model achieved ~99% training accuracy while maintaining strong generalization. Early stopping was applied to prevent overfitting. Model weights were restored from the epoch with the highest validation accuracy. These results indicate that the model effectively distinguishes between the four Alzheimer’s disease stages with high reliability.
Tools & Technologies
Python
TensorFlow / Keras
NumPy
Matplotlib
Google Colab
KaggleHub
Features
Multi-class MRI image classification
Handles class imbalance during training
Model saving and fine-tuning
Function for testing single images with visualization
📁 Project Structure ├── data/ │ ├── NonDemented/ │ ├── VeryMildDemented/ │ ├── MildDemented/ │ └── ModerateDemented/ ├── notebooks/ │ └── Alzheimer_Classification.ipynb ├── saved_model/ ├── README.md
Author
ADIJAT Motunrayo Oyetoke AI / Machine Learning Enthusiast
Future Improvements
Improve recall for minority classes