Machine learning (ML) has been extensively applied in brain imaging studies to aid the diagnosis of psychiatric disorders and the selection of potential biomarkers. Due to the high dimensionality of imaging data and heterogeneous subtypes of psychiatric disorders, the reproducibility of ML results in brain imaging studies has drawn increasing attention. The reproducibility in brain imaging has been primarily examined in terms of prediction accuracy. However, achieving high prediction accuracy and discovering relevant features are two separate but related goals. An important yet under-investigated problem is the reproducibility of feature selection in brain imaging studies. We propose a new metric to quantify the reproducibility of neuroimaging feature selection via bootstrapping. We estimate the reproducibility index (R-index) for each feature as the reciprocal coefficient of variation of absolute mean difference across a larger number of bootstrap samples. We then integrate the R-index in regularized classification models as penalty weight. Reproducible features with a larger R-index are assigned smaller penalty weights and thus are more likely to be selected by our proposed models. Both simulated and multimodal neuroimaging data are used to examine the performance of our proposed models.Results show that our proposed R-index models are effective in separating informative features from noise features. Additionally, the proposed models yield similar or higher prediction accuracy than the standard regularized classification models while further reducing coefficient estimation error. Improvements achieved by the proposed models are essential to advance our understanding of the selected brain imaging features as well as their associations with psychiatric disorders.

Multimorbidity, co-occurrence of two or more chronic conditions, is one of the top priorities in global health research and has emerged as the gold standard approach to study disease accumulation. As aging underlies the development of many chronic conditions, surrogate aging biomarkers are not disease-specific and capture health at the whole person level, having the potential to improve our understanding of multimorbidity. Biological age has been examined in recent years as a surrogate biomarker to capture the process of aging. However, relatively few studies have investigated the relationship between biological age and multimorbidity. More research is needed to quantify biological age using a broad range of biological markers and multimorbidity based on a comprehensive set of chronic conditions. Brain age estimated by neuroimaging data and machine learning models is another surrogate aging biomarker predictive of a wide range of health outcomes. Little is known about the relationship between brain age and multimorbidity. To answer these questions, our study investigates whether elevated biological age and accelerated brain age are associated with an increased risk of multimorbidity using a large dataset from the Midlife in the United States (MIDUS) Refresher study. Ensemble learning is utilized to combine multiple machine learning models to estimate biological age using a comprehensive set of biological markers. Brain age is obtained using convolutional neural networks and neuroimaging data. Our study is the first to examine the relationship between accelerated brain age and multimorbidity and presents the first effort to test whether sex moderates the relationship between these surrogate aging biomarkers and multimorbidity. Furthermore, it is the first attempt to explore how biological age and brain age are related to multimorbidity in mental health. Our findings hold the potential to advance the understanding of the accumulation of physical and mental health conditions, which may contribute to new strategies to improve the treatment of multimorbidity and detection of at-risk individuals.