Hierarchical multi-class Alzheimer's disease diagnostic framework using imaging and clinical features

Area of Science:

• Neuroscience
• Medical Imaging
• Biostatistics

Background:

• Early diagnosis of Alzheimer's disease (AD) is challenging due to its clinical continuum.
• Current diagnostic accuracy for early stages of AD requires improvement.

Purpose of the Study:

• To create a hierarchical multi-class diagnostic framework for normal cognition (NC), early mild cognitive impairment (EMCI), late mild cognitive impairment (LMCI), and AD.
• To investigate the geometric properties of cerebral cortical structures for diagnostic potential.

Main Methods:

• Utilized data from 1,670 participants in the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.
• Extracted four cortical geometric properties (cortical thickness, fractal dimension, gyrification index, sulcus depth) for 148 anatomical structures.
• Integrated imaging features with Mini-Mental State Examination (MMSE) scores and applied AdaBoost algorithm with sparse learning for feature selection.

Main Results:

• Achieved an overall accuracy of 0.877 in the temporal validation set using the developed framework.
• Identified key discriminative imaging features, including cortical thickness and sulcus depth in specific brain regions like the posterior cingulate gyrus and superior temporal sulcus.
• Clinical Dementia Rating (CDR) emerged as a primary clinical variable associated with AD progression.

Conclusions:

• The hierarchical framework effectively combines cognitive variables and neuroimaging features for improved AD diagnosis.
• Surface-based morphometry and geometric analysis of cortical structures show reliability for early AD detection.
• This approach facilitates more accurate clinical diagnosis of Alzheimer's disease.