Histogram-Derived MREPT Connectome: A Distribution-Aware Framework for Brain Network Analysis

Conventional magnetic resonance electrical properties tomography (MREPT) has primarily focused on reconstructing conductivity maps and comparing mean regional values. Such mean-based approaches discard distributional information and overlook network-level alterations. Phase-derived conductivity is also sensitive to noise, motivating the development of a distribution-aware framework that remains robust and biologically interpretable. A histogram-derived MREPT connectome was introduced, representing each ROI by the full conductivity distribution and defining inter-regional edges using Kullback–Leibler (KL) divergence–based similarity. The framework captures differences across entire histograms rather than central tendencies alone, with edge weights anchored to interpretable shape indices such as entropy and interquartile range. The approach avoids reliance on tractography or time-series data. Application to cognitively normal participants demonstrated measurable correspondence with independent structural connectivity, reproducible identification of a stable core subset of regions, and robust edge–shape associations independent of spatial distance. Group-level comparison further revealed network disruption in Alzheimers disease, characterized by reduced connectivity strength and integration together with stronger modular segregation. Regional alterations concentrated in temporal, fusiform, and inferior parietal cortices, accompanied by diminished hub prominence in default-mode territories. The histogram-derived MREPT connectome provides a noise-tolerant, distribution-sensitive, and interpretable network framework capable of capturing clinically relevant disruption and establishing a foundation for reproducible group comparisons and integration with complementary neuroimaging approaches.