AI-driven Classification of Alzheimer's Disease and Mild Cognitive Impairment

Overview

This project employs advanced machine learning algorithms to classify Alzheimer's Disease with high accuracy. By leveraging a dataset of cognitive and biological markers, the code achieves remarkable performance metrics, including: (Accurate measurements are not available until the paper is published!)

• Accuracy: >80%
• Sensitivity: >80%
• Specificity: >80%
• F1 Score: >80%

The implementation is designed for ease of use and reproducibility, enabling researchers and practitioners to effectively diagnose Alzheimer's Disease.

Key Features

• Implements a Multinomial Logistic Regression (MLR) & Multi-Layer Perceptron (MLP)
• Multi-class classification.
• Optimized hyperparameters for high performance.
• Outputs detailed evaluation metrics (accuracy, sensitivity, specificity, F1 score).
• Configurable for different datasets.

Data

Data Source

This project uses data from the Alzheimer's Disease Neuroimaging Initiative (ADNI). ADNI is a longitudinal multicenter study designed to develop clinical, imaging, genetic, and biochemical biomarkers for the early detection and tracking of Alzheimer's disease.

To access ADNI data, you must apply for access through their Data Access Application.

Please note that this repository does not contain any ADNI data due to restrictions on data use agreements.

Our Dataset

We collected a dataset consisting of MRI T1-W, FDG-PET, comorbidities, and demographic information for all subjects from the ADNI database. The dataset template template has been added to the repository as Features-template.xlsx

Data Analysis

The PET and MRI images were preprocessed using Python and SPM12. Data cleaning, outlier handling, and data manipulation were implemented using Python. MLR and MLP models were developed as follows.

Model Summaries

MLR Summary

Layer (type)	Input Shape	Output Shape	Param #	Trainable
LogisticRegression	[8, 1, 237]	[8, 1, 3]	--	True
├─ Linear (linear)	[8, 1, 237]	[8, 1, 3]	714	True

Total Parameters: 714
Trainable Parameters: 714
Non-trainable Parameters: 0
Total Mult-Adds (M): 0.01

Resource	Size (MB)
Input size	0.01
Forward/backward pass size	0.00
Params size	0.00
Estimated Total Size	0.01

---

MLP Summary

Layer (type)	Input Shape	Output Shape	Param #	Trainable
MLP	[8, 1, 237]	[8, 1, 3]	--	True
├─ Linear (fc1)	[8, 1, 237]	[8, 1, 64]	15,232	True
├─ Linear (fc2)	[8, 1, 64]	[8, 1, 64]	4,160	True
├─ Linear (fc3)	[8, 1, 64]	[8, 1, 32]	2,080	True
├─ Linear (fc4)	[8, 1, 32]	[8, 1, 32]	1,056	True
├─ Linear (fc5)	[8, 1, 32]	[8, 1, 3]	99	True

Total Parameters: 22,627
Trainable Parameters: 22,627
Non-trainable Parameters: 0
Total Mult-Adds (M): 0.18

Resource	Size (MB)
Input size	0.01
Forward/backward pass size	0.01
Params size	0.09
Estimated Total Size	0.11

---

This structure is neat, accessible, and easy to edit. Let me know if you’d like to make additional adjustments!

Requirements

To run this project, install the following dependencies:

• Python 3.8+
• Required libraries (install via requirements.txt):

  pip install -r requirements.txt

Usage

Step 1: Clone the Repository

git clone https://github.com/taha-parsayan/Classification-of-Alzheimers-Disease-using-AI-algorithms.git
cd Classification-of-Alzheimers-Disease-using-AI-algorithms

Step 2: Prepare the Data

Place your dataset in the data/ directory. Modify the data loading section in main.py to specify your dataset path and adjust preprocessing steps as needed.

Step 3: Run the Code

Execute the script to train and evaluate the model:

python main.py

Step 4: View Results

The results, including accuracy, sensitivity, specificity, and F1 score, will be displayed in the console and saved to results/.

Code Structure

Classification-of-Alzheimers-Disease-using-AI-algorithms/
│
├── main.py                # Main script for training and evaluation
├── data/                  # Directory for the dataset
├── models/                # Directory for saving trained models
├── results/               # Directory for saving results
├── utils.py               # Utility functions (if applicable)
└── requirements.txt       # Required Python libraries

Performance Metrics

Metric	Value (%)
Accuracy	>80
Sensitivity	>80
Specificity	>80
F1 Score	>80

License

This project is licensed under the MIT License. See the LICENSE file for details.

Author

Taha Parsayan

For questions or collaborations, contact: GitHub Profile