Toward a Multimodal Computer-Aided Diagnostic Tool for Alzheimer's Disease Conversion

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder impacting memory and daily living.
• Mild cognitive impairment (MCI) patients have a high risk of progressing to AD.
• Current diagnostic methods for AD conversion rely on clinical assessment and MRI, with limited automated tools for longitudinal MCI monitoring.

Purpose of the Study:

• To develop an automated computer-assisted diagnosis (CAD) system for predicting AD conversion in MCI patients using longitudinal data.
• To address the gap in automated diagnostic tools for monitoring MCI progression in clinical settings.
• To identify key neuroimaging and clinical features associated with AD conversion.

Main Methods:

• A computationally efficient pre-processing and prediction pipeline leveraging longitudinal T1-weighted MRI, cognitive tests, and demographics.
• A convolutional neural network combining Attention and Inception architectures for pattern recognition.
• Linear registration for efficient image processing between time points.

Main Results:

• The proposed model achieved an Area Under the Curve (AUC) of 84.7% for predicting AD conversion in MCI subjects.
• This performance was significantly better than using cognitive tests and demographics alone (AUC = 80.6%).
• Key predictive features included the thalamus, caudate, planum temporale, and the Rey Auditory Verbal Learning Test.

Conclusions:

• The developed CAD system offers an objective tool for diagnosing AD conversion in MCI patients.
• The method is computationally efficient and can be integrated into clinical workflows.
• Leveraging longitudinal imaging and clinical data enhances prediction accuracy for MCI to AD progression.