Predicting positron emission tomography brain amyloid positivity using interpretable machine learning models with wearable sensor data and lifestyle factors

Abstract

Background: Developing a screening method for identifying individuals at higher risk of elevated brain amyloid burden is important to reduce costs and burden to patients in clinical trials on Alzheimer's disease or the clinical setting. We developed machine learning models using objectively measured lifestyle factors to predict elevated brain amyloid burden on positron emission tomography.

Methods: Our prospective cohort study of non-demented, community-dwelling older adults aged ≥ 65 years was conducted from August 2015 to September 2019 in Usuki, Oita Prefecture, Japan. One hundred and twenty-two individuals with mild cognitive impairment or subjective memory complaints (54 men and 68 women, median age: 75.50 years) wore wearable sensors and completed self-reported questionnaires, cognitive test, and positron emission tomography imaging at baseline. Moreover, 99 individuals in the second year and 61 individuals in the third year were followed up. In total, 282 eligible records with valid wearable sensors, cognitive test results, and amyloid imaging and data on demographic characteristics, living environments, and health behaviors were used in the machine learning models. Amyloid positivity was defined as a standardized uptake value ratio of ≥ 1.4. Models were constructed using kernel support vector machine, Elastic Net, and logistic regression for predicting amyloid positivity. The mean score among 10 times fivefold cross-validation repeats was utilized for evaluation.

Results: In Elastic Net, the mean area under the receiver operating characteristic curve of the model using objectively measured lifestyle factors alone was 0.70, whereas that of the models using wearable sensors in combination with demographic characteristics and health and life environment questionnaires was 0.79. Moreover, 22 variables were common to all machine learning models.

Conclusion: Our machine learning models are useful for predicting elevated brain amyloid burden using readily-available and noninvasive variables without the need to visit a hospital.

Trial registration: This prospective study was conducted in accordance with the Declaration of Helsinki and was approved by the local ethics committee of Oita University Hospital (UMIN000017442). A written informed consent was obtained from all participants. This research was performed based on the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline.

Keywords: Amyloid positivity; Lifestyle factors; Machine learning; Mild cognitive impairment; PiB-PET; Wearable sensor.

Fig. 1

Receiver operating characteristic (ROC) curves in each three machine learning models. Every ROC curve represents the results in the fifth model from the top of 10 times seeds for the fivefold cross validation results of predicting amyloid positivity. The blue line shows the mean ROC for the fivefold cross validation, and the red dot line shows the chance and ROC of each fivefold according to different colors. The gray shadow shows ± 1 standard deviation of the mean ROC. a kernel SVM, b Elastic Net, c logistic regression

Fig. 2

The boxplot of the cross-validated AUCs across all 10 repeats. The flier points are those past the end of the whiskers extending from the box by 1.5 × the inter-quartile range (IQR). a kernel SVM, b Elastic Net, c logistic regression

Fig. 3

The feature importance ranking table extracted in each three machine learning models. The vertical axis labels show the explanatory variables, and the horizontal axis labels depict the feature importance of each explanatory variable. a kernel SVM, b Elastic Net, c logistic regression

Fig. 4

Variables that remained in each of the model. Venn diagram showing the variables that finally remained among the three models and 22 common variables in all three models