Robust Identification of Rich-Club Organization in Weighted and Dense Structural Connectomes

The human brain is a complex network, in which some brain regions, denoted as 'hub' regions, play critically important roles. Some of these hubs are highly interconnected forming a rich-club organization, which has been identified based on the degree metric from structural connectomes constructed using diffusion tensor imaging (DTI)-based fiber tractography. However, given the limitations of DTI, the yielded structural connectomes are largely compromised, possibly affecting the characterization of rich-club organizations. Recent progress in diffusion MRI and fiber tractography now enable more reliable but also very dense structural connectomes to be achieved. However, while the existing rich-club analysis method is based on weighted networks, it is essentially built upon degree metric and, therefore, not suitable for identifying rich-club organizations from such dense networks, as it yields nodes with indistinguishably high degrees. Therefore, we propose a novel method, i.e. Rich-club organization Identification using Combined H-degree and Effective strength to h-degree Ratio (RICHER), to identify rich-club organizations from dense weighted networks. Overall, it is shown that more robust rich-club organizations can be achieved using our proposed framework (i.e., state-of-the-art fiber tractography approaches and our proposed RICHER method) in comparison to the previous method focusing on weighted networks based on degree, i.e., RC-degree. Furthermore, by simulating network attacks in 3 ways, i.e., attack to non-rich-club/non-rich-club edges (NRC2NRC), rich-club/non-rich-club edges (RC2NRC), and rich-club/rich-club edges (RC2RC), brain network damage consequences have been evaluated in terms of global efficiency (GE) reductions. As expected, significant GE reductions have been detected using our proposed framework among conditions, i.e., NRC2NRC < RC2NRC, NRC2NRC < RC2RC and RC2NRC < RC2RC, which however have not been detected otherwise.