# Effective workflow from multimodal MRI data to model-based prediction

## TL;DR

This paper introduces a new framework that uses brain models and MRI data to better predict human behavior and traits.

## Contribution

A novel model-based workflow that improves prediction performance by integrating simulated brain data with machine learning.

## Key findings

- Incorporating simulated data improves prediction performance over empirical features alone.
- The approach was successfully applied to sex classification and prediction of cognition or personality traits.
- Dynamical brain models can capture brain features not easily measured directly.

## Abstract

Predicting human behavior from neuroimaging data remains a complex challenge in neuroscience. To address this, we propose a systematic and multi-faceted framework that incorporates a model-based workflow using dynamical brain models. This approach utilizes multi-modal MRI data for brain modeling and applies the optimized modeling outcome to machine learning. We demonstrate the performance of such an approach through several examples such as sex classification and prediction of cognition or personality traits. We in particular show that incorporating the simulated data into machine learning can significantly improve the prediction performance compared to using empirical features alone. These results suggest considering the output of the dynamical brain models as an additional neuroimaging data modality that complements empirical data by capturing brain features that are difficult to measure directly. The discussed model-based workflow can offer a promising avenue for investigating and understanding inter-individual variability in brain-behavior relationships and enhancing prediction performance in neuroimaging research.

The online version contains supplementary material available at 10.1038/s41598-025-04511-5.