Personalized Diagnosis for Alzheimer's Disease

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Current Alzheimer's Disease (AD) diagnosis methods use general classifiers trained on large, homogeneous datasets.
• Real-world clinical imaging data is complex and heterogeneous due to diverse disease pathology, limiting the effectiveness of standard models.
• Existing methods often fail to achieve expected outcomes in routine clinical practice.

Purpose of the Study:

• To develop a novel personalized model for accurate Alzheimer's Disease diagnosis.
• To address the limitations of general classifiers in heterogeneous clinical imaging data.
• To improve diagnostic accuracy and clinical score estimation at the individual level.

Main Methods:

• Proposed a personalized model for AD diagnosis by customizing subject-specific classifiers.
• Employed iterative reweighting of training data to reveal latent testing data distribution.
• Refined classifiers based on weighted training data and extended the model to a joint classification and regression framework.

Main Results:

• The personalized model demonstrated improved classification and regression accuracy.
• Enhanced performance was observed when applied to Magnetic Resonance Imaging (MRI) data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.
• The joint classification and regression approach improved estimation of diagnosis results and clinical scores.

Conclusions:

• The personalized diagnosis framework shows significant clinical potential for Alzheimer's Disease.
• Subject-specific classification tailored to individual data distributions enhances diagnostic performance.
• This approach offers a promising direction for improving AD diagnosis in clinical settings.