Biomarkers.

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Neuropathological burden in Alzheimer's disease (AD) correlates with clinical symptoms, suggesting in-vivo markers can detect early disease stages.
• Analysis of clinical and neuroimaging data from mild-moderate AD patients aimed to identify the disease in earlier clinical stages.
• A multimodal diagnostic framework was developed using risk factors and MRI brain markers, evaluated via random forests classification.

Purpose of the Study:

• To identify early-stage Alzheimer's disease (AD) using a multimodal diagnostic framework.
• To evaluate the effectiveness of combining neuroimaging features and non-imaging risk factors for AD diagnosis.
• To build and assess a random forest classification model for early AD detection.

Main Methods:

• Utilized the OASIS-3 database with longitudinal data from over 800 individuals (CDR score 0-2).
• Extracted causal MRI features and non-imaging risk factors (demographics, family history, medical history) associated with AD.
• Developed a random forest classification model using 35 identified features to diagnose mild-moderate AD.

Main Results:

• Identified causal relationships between mild-moderate AD and 24 brain regions, including the hippocampus and entorhinal cortex.
• Key risk factors identified include age, APOE ɛ4 carrier status, maternal dementia history, hypertension, and stroke.
• The random forest model achieved 92% balanced accuracy, with 97% sensitivity and 87% specificity for AD detection.

Conclusions:

• Integrating specific MRI features and risk factors shows promise for detecting at-risk and early-stage Alzheimer's disease patients.
• This approach can aid in the early identification of individuals who may benefit from timely intervention.
• The study highlights the potential of multimodal data analysis in improving AD diagnosis.