Correlation Between Brain MRI and Continuous Physiological and Environmental Traits Using 2D Global Descriptors and Multi-Order Image Transforms

MRI is a highly prevalent method of diagnosing neurological conditions and brain anomalies, as well as studying brain connectivity and development. Here we describe a method that applies a large set of global numerical image content descriptors to brain MRI images and correlates the visual content with continuous values. Unlike some other approaches, the proposed method does not attempt to detect discriminating regions, but discriminating image content descriptors that reflect changes in the entire imaged region. Experimental results using a dataset of MRI images acquired from 416 subjects show that the method detected strong correlation between brain MRI images and basic physiological indicators such as age and gender, but also found significant correlation between the brain MRI images and the level of education or socio-economical status. This approach of brain image analysis can be used in population studies for detecting biomarkers, as well as correlation of structure of the brain with continuous physiological, environmental, or behavioral traits. Software and source code for the method are publicly available for free download.