IQ Estimation Using Resting State fMRI and Regression Dynamic Causal Modeling

Intelligence is a multifaceted concept that is generally defined as the ability of comprehending the environment, reasoning, planning, thinking, and learning. The Intelligence Quotient (IQ) is a well-known measure of cognitive abilities associated with intelligence. Recent neuroimaging studies have shown that brain structure and function are related to IQ, which varies from person to person. In this study, we investigated the potential of neuronal effective (directional) connectivity, computed using regression Dynamic Causal Modeling (rDCM) applied to resting-state fMRI data from 100 unrelated subjects in the HCP database, as a source of information for IQ estimation. We used permutation t-test and step-wise wrapper methods to select 24 significant features derived from directional connectivity and applied machine learning algorithms of Support Vector Regression (SVR) and Multi-Layer Perceptron (MLP) with 7-fold cross-validation to predict IQ. The results showed a reasonable IQ estimate, with a mean absolute error (MAE) of about 2 points and a prediction determination coefficient (R2) of about 70 percent. Our findings suggest that neuronal effective connectivity may provide valuable information for IQ estimation and support further investigation into its potential as a biomarker for intelligence.