Network based Identification of Dementia Onsets Using Structural MRI Network Signature

Dementia is one of the leading causes of mortality across the globe, yet, its treatment remains practically elusive. Preventing the later onsets of dementia by treating the early onset stands a better chance to forestall further surge in the cases of dementia around the world. Unfortunately, a lot of issues are associated with the detection of early onset as their clinical symptoms overlap with those of normal aging. Therefore, in this framework, the gray matter tissue probability map (TPM) is extracted from the magnetic resonance imaging (MRI) data of dementia related subjects. Generalized improved multiscale permutation entropy (GIMPE) based networks are formulated on the extracted gray matter TPM of normal control (NC), stable mild cognitive impairment (sMCI), progressive mild cognitive impairment (sMCI) and Alzheimer's disease (AD) subjects. The network disruption of dementia onsets are assessed taking the networks of NC subjects as reference. A technique is developed and validated for the formulation of brain network from connectivity matrix and the formulated network topologies are quantified using graph theory metrics at nodal levels. The topological metrics at nodal levels are statistically analyzed using a non-parametric statistical (Kruskal-Wallis) test to extract the network signatures corresponding to NC, sMCI, pMCI and AD groups. Results show that the proposed framework is potentially viable for the detection and identification of the normal aging as well as the various stages of dementia onset at network level.