Structural differences in adolescent brains can predict alcohol misuse

Abstract

Alcohol misuse during adolescence (AAM) has been associated with disruptive development of adolescent brains. In this longitudinal machine learning (ML) study, we could predict AAM significantly from brain structure (T1-weighted imaging and DTI) with accuracies of 73 - 78% in the IMAGEN dataset (n ~1182). Our results not only show that structural differences in brain can predict AAM, but also suggests that such differences might precede AAM behavior in the data. We predicted ten phenotypes of AAM at age 22 using brain MRI features at ages 14, 19, and 22. Binge drinking was found to be the most predictable phenotype. The most informative brain features were located in the ventricular CSF, and in white matter tracts of the corpus callosum, internal capsule, and brain stem. In the cortex, they were spread across the occipital, frontal, and temporal lobes and in the cingulate cortex. We also experimented with four different ML models and several confound control techniques. Support Vector Machine (SVM) with rbf kernel and Gradient Boosting consistently performed better than the linear models, linear SVM and Logistic Regression. Our study also demonstrates how the choice of the predicted phenotype, ML model, and confound correction technique are all crucial decisions in an explorative ML study analyzing psychiatric disorders with small effect sizes such as AAM.

Machine learning-based prediction of clinical outcomes after first-ever ischemic stroke

“…Due to significant differences in the number and mean age of female and male patients, we balanced the dataset by separating all patients into groups according to sex and age and then randomly selecting patients within these groups until there were no more significant differences (up to p ≤ 0.1). This was necessary to ensure the predictions of our models were not based on an inherent bias in the training data (e.g., women being older on average and thus having worse outcomes) ( 16 ). The patient selection process is shown in Figure 1 and the characteristics of the dataset are described in Table 1 .…”

“…To reduce complexity and potential problems brought on by multiple comparisons, a small set of three ML algorithms were selected. A Support Vector Machine (SVM) with linear kernel (SVM-lin) ( 19 ) and a SVM with radial basis function kernel (SVM-rbf) ( 20 ) were chosen as linear and non-linear models due to their strong performance in previous studies and the ability to directly compare them ( 6 , 16 , 21 ). Similarly, Gradient Boosting (GB) ( 22 ) was chosen as the tree-based classifier due to its superior performance and when compared to other tree-based models ( 23 , 24 ).…”

Uncontrolled eating and sensation-seeking partially explain the prediction of future binge drinking from adolescent brain structure

“…In line with this, using a more comprehensive methodological approach, we recently demonstrated that binge drinking at age 22 can be predicted from structural magnetic resonance imaging (sMRI) features at age 14 with an average balanced accuracy of more than 70%. Of note, the number of lifetime binge drinking occasions was the best predictable among multiple phenotypes of alcohol misuse (11).…”

“…Achieving this necessitates an understanding of the neuro-psychological predispositions driving risky alcohol consumption during adolescence (8). Several recent studies suggest that differences in adolescent brain structure can predict binge drinking (9)(10)(11). Longitudinal multi-site projects, such as the IMAGEN study (12), provide a unique opportunity to identify neural predictors of binge drinking in a large-scale sample of adolescents (8).…”

“…The accurate prospective prediction of a behavior as complex as binge drinking based on structural brain features (11) inevitably evokes one question: Which role do previously unconsidered variables play? Such variables can, i.a., be alcohol-related measures (e. g. drinking motives, consumption frequency), other binge-and addiction-related variables (e. g. uncontrolled eating, nicotine dependence), psychiatric comorbidities (e. g. depression, anxiety), personality (e. g. extraversion, impulsivity) or social factors (e. g. bullying, prosocial behavior).…”