An efficient approach for differentiating Alzheimer's disease from normal elderly based on multicenter MRI using...

Area of Science:

• Neuroimaging
• Machine Learning
• Biomedical Engineering

Background:

• Machine learning and structural MRI markers improve Alzheimer's disease (AD) detection accuracy.
• Current methods using non-linear transformations face challenges with parameter dependency and registration errors.
• Anatomical features like cortical volume, shape, and thickness show discriminative potential.

Purpose of the Study:

• To develop a simple, efficient method for extracting disease-related anatomical features from MRI data.
• To enable initial stratification of heterogeneous patient populations in clinical settings.
• To characterize Alzheimer's disease (AD)-specific anatomical features using invariant properties.

Main Methods:

• Employed gray-level invariant features extracted from linearly transformed structural MRI.
• Utilized disease-specific spatial masking and linear registration for anatomical feature capture.
• Implemented a two-step feature selection: statistic-based selection followed by knowledge-based ROI analysis.

Main Results:

• The proposed method effectively differentiated Alzheimer's disease (AD) and mild cognitive impairment (MCI) from normal elderly controls (NC).
• Statistic-based feature selection and knowledge-based masks efficiently captured relevant brain anatomical features.
• The approach demonstrated promising performance using a support vector machine classifier.

Conclusions:

• The developed method offers a robust solution for efficient brain anatomical feature extraction.
• This approach facilitates rapid data stratification for heterogeneous clinical populations.
• It can be integrated with more complex analyses, such as non-linear transformations, for deeper insights.