Morphological Factor Estimation via High-Dimensional Reduction: Prediction of MCI Conversion to Probable AD

Area of Science:

• Neuroimaging
• Medical Diagnostics
• Biostatistics

Background:

• Alzheimer's disease (AD) diagnosis relies on clinical assessment and neuroimaging.
• Current neuroimaging methods for AD often lack quantitative measures of brain morphology.
• Developing objective biomarkers for early AD detection and progression prediction is crucial.

Purpose of the Study:

• To introduce a novel morphological factor estimation method using structural MRI.
• To evaluate the efficacy of this method in discriminating Alzheimer's disease (AD) from controls.
• To assess the predictive power of the method for Mild Cognitive Impairment (MCI) to AD conversion.

Main Methods:

• A reference MRI feature eigenspace was created using intensity and local volume data from healthy subjects.
• Structural MRI data from probable AD, control (CTRL), and MCI subjects were projected into this eigenspace.
• High-dimensional discriminant functions were extracted, and a single morphological factor was calculated using an inverse-squared model.

Main Results:

• The inverse-squared model demonstrated superiority over other tested models.
• The method achieved 90% accuracy in discriminating probable AD from CTRL.
• The method predicted MCI to probable AD conversion with 79.4% accuracy.

Conclusions:

• The proposed morphological factor estimation from structural MRI is a promising tool for disease state evaluation in AD.
• This quantitative MRI-based method offers high accuracy in AD diagnosis and MCI conversion prediction.
• The findings support the potential clinical utility of this novel neuroimaging biomarker.