Landmark-Based Alzheimer's Disease Diagnosis Using Longitudinal Structural MR Images.

Area of Science:

• Neuroimaging
• Medical Diagnostics
• Machine Learning

Background:

• Alzheimer's disease (AD) diagnosis relies on accurate analysis of structural Magnetic Resonance Imaging (sMRI).
• Longitudinal studies are crucial for tracking disease progression but are sensitive to scan inconsistencies.
• Current methods often require complex preprocessing like nonlinear registration or tissue segmentation.

Purpose of the Study:

• To develop a robust and efficient feature extraction method for AD diagnosis using longitudinal sMRI.
• To overcome limitations of nonlinear registration and tissue segmentation in clinical application.
• To improve the accuracy of distinguishing Alzheimer's disease (AD) and mild cognitive impairment (MCI) from healthy controls (HCs).

Main Methods:

• Automatic discovery of discriminative brain landmarks.
• Fast landmark detection for efficient localization in testing images.
• Extraction of high-level statistical spatial and contextual longitudinal features based on landmarks.
• Classification using a linear Support Vector Machine (SVM).

Main Results:

• The proposed method achieves competitive classification accuracies for AD vs. HC and MCI vs. HC.
• The landmark-based approach demonstrates robustness to inconsistencies in longitudinal scans.
• The method offers promising computational efficiency, reducing diagnostic time.

Conclusions:

• Landmark-based feature extraction offers an effective and efficient alternative for AD diagnosis using longitudinal sMRI.
• The method's independence from nonlinear registration and segmentation simplifies clinical application.
• This technique shows potential for early and accurate detection of AD and MCI.