2DSDNN: A Novel Approach for Alzheimer's Disease Classification.

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Alzheimer's Disease (AD) is a leading cause of dementia in the elderly, presenting diagnostic challenges.
• Accurate staging of AD is crucial for effective patient management and treatment.
• Neurological imaging combined with machine learning shows promise for early AD identification.

Purpose of the Study:

• To develop and validate a deep learning model for precise classification of Alzheimer's Disease (AD) stages using MRI data.
• To improve early detection and differentiation of AD, Mild Cognitive Impairment (MCI), and Cognitively Normal (CN) individuals.
• To enhance the performance of AD classification through advanced image segmentation techniques.

Main Methods:

• A skull-stripping algorithm using thresholding and morphological manipulations was applied to T1- and T2-weighted MRI scans.
• A 2D Sequential Deep Learning Neural Network (2DSDNN) was developed and trained on 1,044 MRI scans from the ADNI dataset.
• The model was evaluated for binary (AD vs. CN, AD vs. MCI) and multiclass (AD vs. MCI vs. CN) classification performance.

Main Results:

• The 2DSDNN model achieved high accuracy across all classification tasks, including 98.03% for AD vs. CN and 97.20% for AD vs. MCI vs. CN.
• Excellent performance metrics were reported, with AUC values reaching up to 98.20% for AD vs. CN classification.
• The model demonstrated superior sensitivity, specificity, precision, and F1-scores in distinguishing between AD, MCI, and CN stages.

Conclusions:

• The proposed 2DSDNN model significantly enhances the performance of binary and multiclass AD classification compared to existing methods.
• The integrated skull-stripping algorithm improves MRI image segmentation, leading to more accurate AD staging.
• This deep learning approach offers a powerful tool for efficient and early detection of Alzheimer's Disease.