Biomarkers.

Area of Science:

• Neurology
• Artificial Intelligence
• Biomarker Discovery

Background:

• Identifying individuals with elevated brain amyloid burden is crucial for Alzheimer's disease (AD) clinical trials and patient care.
• Current screening methods can be costly and burdensome for patients.
• Developing noninvasive, cost-effective screening tools is a priority in AD research.

Purpose of the Study:

• To develop and validate machine learning models for predicting elevated brain amyloid burden.
• To assess the efficacy of lifestyle factors and wearable sensor data in predicting amyloid burden.
• To provide a noninvasive screening tool for individuals at risk of AD.

Main Methods:

• Prospective cohort study (USUKI cohort) data from 118 individuals with mild cognitive impairment (MCI) or subjective memory complaints.
• Development of three machine learning models: kernel support vector machine, Elastic Net, and logistic regression.
• Utilized objectively measured lifestyle factors, demographic characteristics, and wearable sensor data.

Main Results:

• The Elastic Net model incorporating lifestyle factors, demographic data, and wearable sensors achieved an area under the receiver operating characteristic curve of 0.79.
• Models using lifestyle factors alone had a mean area under the curve of 0.70.
• Twenty-two variables were consistently identified across all developed machine learning models.

Conclusions:

• Machine learning models effectively predict elevated brain amyloid burden in MCI patients.
• The models leverage readily available, noninvasive variables, reducing the need for hospital visits.
• This approach offers a promising tool for early detection and risk stratification in Alzheimer's disease.