Improving Alzheimer Diagnoses With An Interpretable Deep Learning Framework: Including Neuropsychiatric Symptoms

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder.
• Neuropsychiatric symptoms (NPS) are common in AD and linked to its severity.
• Current research methods for linking NPS and AD severity lack intuitive approaches.

Purpose of the Study:

• To develop a hybrid deep learning framework for accurate Alzheimer's disease diagnosis.
• To integrate multimodal data including neuroimaging and clinical information.
• To investigate the role of specific neuropsychiatric symptoms in AD diagnosis.

Main Methods:

• A hybrid deep learning framework combining a 3D convolutional neural network (CNN) and Principal Component Analysis (PCA).
• Utilized multimodal inputs: structural MRI, behavioral scores, age, and gender.
• Employed SHapley Additive exPlanations (SHAP) for interpreting model decisions.

Main Results:

• Achieved high accuracy (0.91) and Area Under the Curve (0.97) in classifying AD from cognitively normal individuals.
• Demonstrated the model's effectiveness in integrating imaging and non-imaging data.
• Identified apathy as a significant neuropsychiatric symptom contributing to AD diagnosis.

Conclusions:

• The proposed hybrid deep learning framework offers a robust method for Alzheimer's disease diagnosis.
• Apathy emerges as a crucial factor for consideration in AD diagnosis, clinical trials, and future research.
• Multimodal data integration significantly enhances diagnostic accuracy for Alzheimer's disease.