Evidence for embracing normative modeling

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Developmental Neuroscience

Background:

• Normative modeling provides a powerful framework for interpreting individual brain data within a developmental context.
• Existing repositories offer valuable normative models but require expansion to cover diverse brain metrics and user accessibility.

Purpose of the Study:

• To expand the normative model repository with lifespan trajectories of brain structural surface area and functional connectivity.
• To develop an updated online platform for seamless transfer of normative models to new datasets.
• To demonstrate the utility of normative modeling features against raw data in neuroimaging analysis.

Main Methods:

• Incorporated normative models for brain structural surface area and functional connectivity using Yeo-17 and Smith-10 atlases.
• Developed an updated online platform for model transferability.
• Conducted head-to-head comparisons using mass univariate group difference testing, classification, and regression tasks.

Main Results:

• Normative modeling features demonstrated superior performance across all benchmarking tasks compared to raw data.
• Statistically significant advantages were most pronounced in group difference testing (schizophrenia vs. control) and classification tasks.
• The updated models and platform facilitate more robust and interpretable neuroimaging analyses.

Conclusions:

• Expanded normative models and an accessible platform enhance the application of normative modeling in neuroimaging.
• Normative modeling offers a significant advantage for detecting group differences, classifying conditions, and predicting cognitive abilities.
• These resources aim to promote wider adoption of normative modeling within the research community.