A Robust Deep Model for Improved Classification of AD/MCI Patients

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Accurate classification of Alzheimer's disease (AD) and mild cognitive impairment (MCI) is crucial for patient outcomes.
• Noninvasive imaging biomarkers are highly sought for early AD diagnosis.

Purpose of the Study:

• To develop a robust deep learning system for classifying AD and MCI progression stages using MRI and PET scans.
• To evaluate the effectiveness of the dropout technique in enhancing deep learning model performance for AD diagnosis.

Main Methods:

• A deep learning framework incorporating dropout, stability selection, adaptive learning rates, and multitask learning was developed.
• The system was applied to the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset for AD and MCI conversion diagnosis.

Main Results:

• The dropout technique demonstrated significant effectiveness, improving classification accuracies by an average of 5.9% compared to classical deep learning methods.
• Experimental results validated the proposed deep learning system's performance in differentiating AD and MCI stages.

Conclusions:

• The proposed deep learning system, particularly with the dropout technique, offers a promising approach for accurate, noninvasive AD and MCI diagnosis.
• This method has the potential to aid in early intervention and improve the quality of life for patients with neurodegenerative diseases.