Multi-task Learning of Structural MRI for Multi-site Classification

Abstract

With the advent of Big Data Imaging Analytics applied to neuroimaging, data from multiple sites need to be pooled into larger samples. However, heterogeneity across different scanners, protocols, and populations renders the task of finding underlying disease signatures challenging. In this chapter, three structural MRI datasets of schizophrenia were collected from different imaging sites. A multi-task learning method was developed to simultaneously learn the site-specific and site-shared features from the multi-site data, which were then used to discriminate schizophrenic patients from normal controls. Experiments show that classification accuracy of multi-site data by using multi-task feature learning outperformed that of single-site data and pooled data and also outperformed other comparison methods. The results indicate that the proposed multi-task learning method is robust in finding consistent and reliable structural brain abnormalities associated with schizophrenia across different sites, in the presence of multiple sources of heterogeneity.