Input Agnostic Deep Learning for Alzheimer's Disease Classification Using Multimodal MRI Images

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Alzheimer's disease (AD) is a progressive neurological disorder impacting memory and cognitive functions.
• Machine learning advancements and accessible medical data have spurred research in AD diagnosis.
• Current multi-modal approaches often require specific data types for accurate classification.

Purpose of the Study:

• To develop and evaluate a multi-modal deep learning model for classifying normal cognition, mild cognitive impairment, and AD.
• To introduce an input-agnostic architecture capable of diagnosing AD using either structural MRI (sMRI) or diffusion tensor imaging (DTI) scans.
• To compare the performance of the proposed model against conventional multi-modal methods.

Main Methods:

• Utilized structural MRI (sMRI) and diffusion tensor imaging (DTI) data from the OASIS-3 dataset.
• Implemented a multi-modal deep learning network incorporating both sMRI and DTI.
• Developed an input-agnostic architecture allowing flexible data input (sMRI or DTI).

Main Results:

• The input-agnostic model demonstrated high diagnostic accuracy.
• Achieved 0.96 accuracy when utilizing both sMRI and DTI scans as input.
• The proposed method offers a flexible alternative to previous multi-modal machine learning techniques.

Conclusions:

• The developed input-agnostic deep learning model shows significant promise for accurate Alzheimer's disease diagnosis.
• This approach enhances diagnostic flexibility by accommodating single or combined imaging modalities.
• Further research can leverage this model for earlier and more precise detection of cognitive decline.