Predicting Skill-Based Task Performance and Learning with fMRI Motor and Subcortical Network Connectivity

摘要

Procedural learning is the process of skill acquisition that is regulated by the basal ganglia, and this learning becomes automated over time through cortico-striatal and cortico-cortical connectivity. In the current study, we use a common machine learning regression technique to investigate how fMRI network connectivity in the subcortical and motor networks are able to predict initial performance and traininginduced improvement in a skill-based cognitive training game, Space Fortress, and how these predictions interact with the strategy the trainees were given during training. To explore the reliability and validity of our findings, we use a range of regression lambda values, sizes of model complexity, and connectivity measurements.