Abstract:

The neuroimaging researches are moving towards a translational neuroscience era that is characterized by the use of multivariate predictive modeling, which is distinct from traditional univariate brain mapping studies. These studies maintain a focus on decoding individual differences in a continuously behavioural phenotype from neuroimaging data using regression-based methods, with an ultimate goal of identifying reliable and objective neuromarkers that can aid in clinical practice at the individual level. The machine learning algorithms-based predictive modeling has been successfully applied in the prediction of multiple important behavioural aspects including cognitive abilities, symptom severity for psychiatric patients, personality, and emotion feeling. The methods used for predictive modeling primarily involve multiple linear regression, least absolute shrinkage and selection operator regression, elastic net, ridge regression, support vector regression, relevance vector regression and partial least square regression. Promisingly, studies that are performed large neuroimaging dataset with rigorously external validation focusing on predicting future behavioural outcomes are encouraged. Moreover, multimodal data can be leveraged to extract the complementary information for investigating individual differences. Overall, predictive modelling opens up novel opportunities to better understand the neurobiological substrates of cognitive abilities and brain disorders.

Key words: Individualized prediction, Translational neuroscience, Magnetic resonance imaging, Schizophrenia, Machine learning