Normative modelling establishes a reference distribution of a brain measure — cortical thickness, subcortical volume, functional connectivity — across a healthy population, while accounting for covariates such as age and sex. An individual patient is then scored by how far they deviate from that reference.
The appeal is that it does not assume patients form a homogeneous group. Case-control comparisons ask whether a patient group differs from controls on average, which dissolves exactly the individual variation that matters clinically. Normative models give a per-subject deviation instead, which is what you need when the same diagnosis produces very different brain phenotypes across people.
Think of it as the neuroimaging analogue of a paediatric growth chart: the value of the chart is not the average curve, it is being able to say where one child sits on it.
Why it appears in this program¶
Both Parkinson’s disease projects use it, for the same reason — neuropsychiatric symptoms in PD vary enormously between individuals, and the clinically useful question is about a particular patient rather than about the group.
CNS Projects¶
Julia-Katharina Pfarr — adapting the pre-estimated Rutherford et al. lifespan model (82 sites, N=58,836, ages 2–100) to new sites and testing it on PD patients, to predict neuropsychiatric symptoms from scans that are already part of routine care
Johanna Bayer — a founding researcher in the field, now developing federated and longitudinal normative models to characterise individual trajectories of brain change across the lifespan, and lead developer of the velocity centiles methodology
Getting started¶
The PCNtoolkit ecosystem is the standard implementation, and Johanna is an active contributor to it.
Structural MRI — cortical thickness and subcortical volume are the usual inputs
Parkinson’s Disease — the topic hub
Aging and Cognitive Decline — separating normal lifespan variation from pathology is the same problem in a different guise