Biomarker Detection from fMRI-Based Complete Functional Connectivity Networks

The comprehensive set of neuronal connections of the human brain, which is known as the human connectomes, have been widely used in the research of neurological and neurodevelopmental disorders. Functional Magnetic Resonance Imaging (fMRI) has facilitated the research by capturing functional activities of the neurons. fMRI-based functional connectomes are used to extract the complete functional connectivity networks, which are edge-weighted complete graphs. In functional connectivity networks, each node represents one brain region or Region of Interest (ROI), and each edge-weight represents the functional similarity of the adjacent ROIs. In order to leverage graph mining methodologies, these complete graphs are often made sparse by applying thresholds on edge-weights. This approach can result in losing discriminative information, while addressing the issue of biomarkers detection, that is, finding discriminative ROIs and connections, given the data of healthy and disabled population. In this work, we present a framework for representing the complete functional connectivity networks in a threshold-free manner and finding the biomarkers by using feature selection algorithms. Additionally, for computing the impact of the biomarkers on the healthy and disabled subjects, we apply tensor decomposition. Experiments on a fMRI dataset of neurodevelopmental reading disabilities show the highly interpretable nature of our approach in finding the biomarkers of the disease.