Radiomics and connectomics are two methodologies widely used in the field of neuroimaging. However, to date, very few studies have attempted to combine these two approaches. This study aimed to evaluate a combined approach based on radiomics and connectomics to study the effect of aging and sex on the brain. Ninety-seven healthy participants underwent both Tl-weighted and resting-state fMRI (rs-fMRI). From Tl-weighted images, the following first order radiomic statistical features were extracted from grey matter: 10th percentile, kurtosis, minimum, range, skewness, and uniformity. From rs-fMRI time-series the following global network metrics were computed from a graph theory model applied to the brain using a cortical parcellation: assortativity, global efficiency, transitivity, modularity, and characteristic path length. A multiple linear regression model was applied to preliminarily assess the predictive power of both radiomic and connectomic features on brain aging. Then, these features were used to train three machine learning algorithms (decision tree, support vector machine, and logistic regression) and to assess their predictive power to discriminate age and sex categories. Results suggest that the proposed combination of radiomic and connectomic features could be useful for characterizing aging, whereas no significant improvements were achieved for sex. Future investigations on larger samples, involving also sub-cortical structures, as well as other types of both radiomic and connectomic features, are needed to confirm the actual potential of the proposed “radioconnectomics” approach in the study of brain aging.
