Predicting intelligence from fMRI data of the human brain in a few minutes of scan time

Abstract

A number of recent studies have investigated machine learning techniques for predicting individual behaviour from fMRI. Even though encouraging results have been obtained, excessive scan times – especially in resting state fMRI – are a limiting factor.Here we present a new machine learning algorithm for predicting individual behaviour of healthy human subjects using both resting state (rsfMRI) as well as task-based fMRI (tfMRI). It achieves dimensionality reduction via ensemble learning and partial least squares regression rather than via brain parcellations or ICA decompositions. In addition, it introduces Ricci-Forman curvature as a novel type of edge weight.As a proof of concept, we focus on predicting fluid, crystallized and general intelligence scores. In a cohort of 390 unrelated test subjects of the Human Connectome Project, we found correlations between the observed and the predicted general intelligence of more than 50 percent in tfMRI, and of around 59 percent (R2 ≈ 0.29) when results from two tasks are combined. We compare these results against a benchmark of existing methods that produced correlations below 50 percent in both rsfMRI and tfMRI. We conclude that with novel machine learning techniques applied to tfMRI it is possible to obtain significantly better prediction accuracies at a fraction of the scan time.

Statistical quantification of confounding bias in machine learning models

“…In the case of the HCP dataset, the statistically significant age bias of the “raw” model for predicting fluid intelligence is in line with previous findings [ 18 , 19 ] and could likely exaggerate to a serious bias when testing the model on data of participants outside of the—relatively narrow—age range of the HCP sample. In this case, the bias would likely significantly harm the out-of-sample generalizability of this model.…”

“…Functional connectivity data from the Human Connectome Project [ 55 ] (HCP) were used to build predictive models of fluid intelligence ( G f ) and to test for the previously discussed confounding effect of age [ 18 , 19 ] and, additionally, the—somewhat underdiscussed—batch-like effect of acquisition date of the data within the course of the data acquisition process.…”

“…Depending on the research question, several demographic and psychometric variables or the time of day of the data acquisition [ 15 ] can emerge as confounders. As a characteristic example, models trained to predict intelligence [ 16 , 17 ] might provide a statistically significant predictive performance by picking up solely on age-related variance [ 18 , 19 ]. Moreover, various types of systematic sampling bias, as well as stochastic group differences in the training sample, can result in confounded models (e.g., racially biased machine learning models [ 6 , 20 , 21 ]).…”

Discrete Ricci curvatures capture age-related changes in human brain functional connectivity networks

“…FRC and ORC have also been applied to brain networks of healthy populations. Specifically, Lohmann et al [42] applied FRC to predict the intelligence of healthy human subjects using fMRI data, and Farooq et al [36] applied ORC to identify changes in brain structural connectivity during healthy aging. However, no previous study has assessed the ability of discrete Ricci curvatures to characterize FCN alterations during healthy aging.…”

Graph Ricci Curvatures Reveal Atypical Functional Connectivity in Autism Spectrum Disorder

“…Recently, Chatterjee et al [48] used a version of FRC to determine the changes in brain functional connectivity related to attention deficit hyperactivity disorder (ADHD). Additionally, FRC has been used to analyze task-based fMRI data [49] as well as to predict intelligence of healthy human subjects [50]. Most of these studies have also contrasted graph Ricci curvatures with standard network measures such as clustering coefficient and node betweenness centrality, and showed that graph Ricci curvatures can provide new information about brain connectivity organization.…”