DCA-Enhanced Alzheimer's detection with shearlet and deep learning integration.

Area of Science:

• Neuroimaging and Artificial Intelligence
• Machine Learning for Medical Diagnosis
• Neurodegenerative Disease Research

Background:

• Alzheimer's dementia (AD) is a progressive neurodegenerative disorder impacting cognitive function and quality of life.
• Current treatments manage AD symptoms, but a cure is unavailable; early diagnosis is critical for effective intervention.
• Machine learning (ML) and deep learning (DL) offer promising avenues for early AD detection through advanced data analysis.

Purpose of the Study:

• To develop a robust and computationally efficient model for accurate early-stage Alzheimer's dementia diagnosis.
• To leverage magnetic resonance imaging (MRI) data for identifying subtle changes indicative of early AD.
• To overcome challenges in MRI data analysis, including high dimensionality and limited sample sizes.

Main Methods:

• Utilized a dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI) comprising 200 patients across different cognitive statuses.
• Employed deep learning models (SqueezeNet-v1.1, MobileNet-v2, Xception, Inception-v3) and 3D shearlet transform for feature extraction from MRI data.
• Integrated deep and shearlet-based features using discriminant correlation analysis (DCA) and classified using Support Vector Machines (SVMs) and Decision Tree Baggers (DTBs).

Main Results:

• Achieved high classification accuracies for early AD detection across different time points: 94.46% (18 months), 92.97% (12 months), and 95.44% (at stable diagnosis).
• The proposed feature representation and classification model demonstrated significant efficacy in identifying AD from MRI scans.
• The integrated approach successfully addressed challenges related to MRI data dimensionality and variability.

Conclusions:

• The developed ML/DL model shows strong potential for accurate and early diagnosis of Alzheimer's dementia using MRI.
• This technology can facilitate timely therapeutic interventions, potentially improving patient outcomes and disease management.
• Further research can refine these methods for broader clinical application in neurodegenerative disease diagnostics.