Epicenter mapping of cortical morphological dedifferentiation in type 2 diabetes mellitus using morphometric inverse divergence.

BACKGROUND

Type 2 diabetes mellitus (T2DM) is associated with widespread cortical alterations and cognitive decline. However, the macroscopic organizational principles of T2DM-associated cortical degradation remain poorly understood. This study investigated whether T2DM-related cortical morphological dedifferentiation follows specific topological network principles.

METHODS

We analyzed structural MRI data from 49 T2DM patients and 49 matched healthy controls (HCs), representing an aging-relevant middle-aged and older-adult cohort. Individualized cortical morphometric similarity networks were constructed using the morphometric inverse divergence (MIND) method. Integrating these networks with normative functional and structural connectomes, we tested classic hub vulnerability, mapped candidate disease-specific epicenters, and evaluated structure-function alignment.

RESULTS

T2DM patients exhibited cortical dedifferentiation, particularly within the primary visual and dorsal attention networks. Spatial permutation testing revealed this distribution did not conform to generic hub vulnerability or uniform local spreading models. Instead, it preferentially aligned with cross-modal candidate epicenters, primarily in the superior parietal, superior frontal, and lingual regions. Crucially, patient-tailored modeling demonstrated a distinct structure-function divergence: individualized dedifferentiation maps correlated significantly with structural, but not functional, epicenter connectivity profiles. Additionally, regional disease duration effects correlated negatively with the structural epicenter profile, tentatively reflecting a ceiling effect in core sensory regions.

CONCLUSION

The group-level spatial map showed exploratory correspondence with selected normative functional connectivity (FC)- and structural connectivity (SC)-based candidate profiles. Individualized profiles showed greater cross- sectional correspondence with the selected SC-based profile than with the FC-based profile, providing a network-informed description of diabetic brain alterations.