XNeuro: An Alzheimer's Disease Diagnosis Framework Using Multi-modal Learning

This project develops a multimodal model to assess Alzheimer's disease using numerical data and MRI images.
The model classifies records into four categories: Non-Demented, Mild Dementia, Moderate Dementia, and Severe Dementia.It integrates diverse data sources to enhance diagnostic accuracy and reliability.

Datasets

OASIS Dataset: Provided by the Washington University Alzheimer’s Disease Research Center.

• Numerical Data: Includes features like ID, Gender, Dominant Hand, Age, Education Level, Socioeconomic Status, MMSE, eTIV, nWBV, and CDR.
• Image Data: MRI Images

Model Architecture

• Numerical Data: Random Forest with XGboost.
• MRI Images: Convolutional Neural Network (CNN).
• Integration: Late fusion of both models for final classification.

Model Comparision for Numerical Data Algorithms:

Algorithm:

Algorithm	Accuracy	Precision	Recall	F1-Score	Loss	Cohen Kappa Score	Matthews Corr. Coeff.	Hamming Loss	Weighted Jaccard
K-Nearest Neighbors	0.791	0.82	0.79	0.80	2.140	0.672	0.674	0.209	0.705
Support Vector Machine	0.896	0.88	0.90	0.89	0.278	0.831	0.832	0.104	0.837
Logistic Regression	0.945	0.95	0.95	0.94	0.178	0.912	0.913	0.055	0.904
Random Forest	0.995	9.98	9.98	9.96	0.135	0.992	0.992	0.005	0.991

Performance-wise order of the models

1. Random Forest
2. Logistic Regression
3. SVM
4. KNN

Booster:

Algorithm	Accuracy	Precision	Recall	F1-Score	Loss	Cohen Kappa Score	Matthews Corr. Coeff.	Hamming Loss	Weighted Jaccard
AdaBoost + Random Forest	0.9502	0.9156	0.9502	0.9298	0.1031	0.9192	0.9230	0.0498	0.9156
Gradient Boosting + Random Forest	0.9851	0.9853	0.9851	0.9851	0.0544	0.9759	0.9760	0.0149	0.9706
Random Forest + XGBoost	0.9950	0.9955	0.9950	0.9951	0.0659	0.9920	0.9920	0.0050	0.9905

Performance-wise order of the Boosters

1. XGBoost
2. Gradient Boosting
3. AdaBoost

CNN Performance Comparison

Model	Final Training Loss	Final Validation Loss	Final Training Accuracy	Final Validation Accuracy
Xception	0.0056	0.0071	0.9998	0.9951
MobileNetV2	0.0247	0.0233	0.9913	0.9951
ResNet50	0.0534	0.0552	0.9962	0.9851
InceptionV3	0.0542	0.0947	0.9912	0.9851
CNN1	0.4633	0.4099	0.8063	0.9303
CNN2	1.1722	1.2789	0.4931	0.4876