Quantitative neuroimaging meets normative modelling: the last mile for precision medicine applications

Quantitative neuroimaging techniques have proven to be powerful tools for studying normal brain physiology and investigating the structural and functional alterations associated with different brain disorders. Despite their huge potential, their contribution to the clinic has largely been limited to supporting (when possible) more standardised neuroradiological imaging procedures. Traditionally, quantitative neuroimaging research relies on cross-sectional analyses, where groups of patients sharing similar diagnostic labels are compared to matched groups of controls. However, this approach may overlook important variations among individuals, particularly in the context of disease diagnosis and treatment response. In this respect, normative modelling (NM) represents a promising innovation for the field of neuroimaging, mapping brain imaging anomalies at the individual level, thus facilitating the shift towards personalised applications. Instead of treating all individuals as part of a homogeneous population, NM describes the variance of imaging phenotypes within a reference population and exploits it to identify subject-specific measures of disease states via the quantification of individual deviations from the normative expected model. In this paper, we showcase the latest research on NM in neuroimaging to investigate neurological and neuropsychiatric disorders. We lead the readers through the key steps of implementing the NM framework, starting from the normative model selection to model estimation, evaluation, and application. Finally, we highlight how the NM is applied across various neuroimaging techniques to pinpoint biomarkers of functional and structural brain alterations.