Alzheimer's diagnosis using deep learning in segmenting and classifying 3D brain MR images.

Area of Science:

• Medical Image Analysis
• Computational Neuroscience
• Artificial Intelligence in Healthcare

Background:

• Dementia, including Alzheimer's Disease (AD), affects millions globally, with significant cognitive and behavioral symptoms.
• Current diagnostic methods for dementia lack standardization, hindering timely and effective treatment.
• Computational diagnosis using brain Magnetic Resonance Imaging (MRI) is crucial for early Alzheimer's Disease detection.

Purpose of the Study:

• To present a novel computational method for diagnosing Alzheimer's Disease (AD) from 3D brain MRI scans.
• To enhance early diagnosis of AD by improving accuracy in image segmentation and classification.
• To leverage deep learning techniques for more effective Alzheimer's Disease detection.

Main Methods:

• A two-phase deep learning approach: I) segmentation and II) classification of brain tissues.
• Segmentation utilizes a combined Gaussian Mixture Model (GMM) and Convolutional Neural Network (CNN).
• Classification employs a hybrid model integrating Extreme Gradient Boosting (XGBoost) and Support Vector Machine (SVM) on segmented tissues.

Main Results:

• The proposed method achieved a Dice score of 0.96 for segmentation across two datasets (AD-86 and AD-126).
• Classification accuracy reached 0.88 on the AD-86 dataset and 0.80 on the AD-126 dataset.
• The combined XGBoost and SVM model demonstrated improved classification performance.

Conclusions:

• Deep learning excels in medical image segmentation and feature extraction for AD diagnosis.
• The integration of XGBoost and SVM significantly enhances diagnostic accuracy for Alzheimer's Disease.
• The developed computational method shows promise for supporting early and accurate AD detection from brain MRIs.