Visualization and Quantification of Differences in Interaction Strength of Sensory and Motor Networks in the Human Brain using Differential Correlation Analysis and Graph Theory

• ¹ Houston Methodist Research Institute, Neurosurgery, United States
• ² Houston Methodist Research Institute, Translational Imaging, United States
• ³ Houston Methodist Research Institute, Radiology, United States
• ⁴ Houston Methodist Hospital, Neurology, United States

Purpose: To compare the interaction strength of sensory and motor networks between healthy controls and subjects diagnosed with complex partial seizures (CPS) when initiating a willed activity using a new graph-theoretical analysis approach Methods: Functional magnetic resonance images (fMRI) (TR=1300 ms) were obtained during the performance of a visual presentation of 10 faces (10 sec, interspersed by green background, 60 sec). Subjects (9 healthy controls (HC), 6 male, average age: 34.7, 5 CPS subjects, male, average age: 28.4) were instructed to squeeze a ball if a face was perceived as unpleasant. Interictal spikes indicative of anomalous brain activity were identified with simultaneous EEG recordings during the fMRI examinations of the CPS subjects. For each subject, a difference correlation map was calculated using a BOLD signal time course from the primary visual cortex (sensory input) and one from the motor cortex (motor output). Networks using a 2D spring-embedded design and using the 3D anatomical Talairach space with correlation coefficients of signal time courses as edge weights were created. Statistical significance between the networks for the two groups was investigated for selected network parameters. 3D printing technology was used to create replica of computer-generated 3D networks. Results: On average, visual cortex, cuneus, precuneus, dorsal lateral prefrontal cortex (DLPC), cingulate gyrus, parahippocampal gyrus, thalamus, lentiform nucleus, caudate and substantia nigra exhibited stronger correlation with the sensory (input) time course, while postcentral gyrus, supplementary motor area, precentral gyrus, insula and regions in the cerebellum showed stronger correlation with the motor (output) time course. Graph-theoretical analysis showed a larger cluster coefficient in the CPS group (p-value<0.05), as well as larger diameter, average path length and larger first eigenvalue (p-value <0.17). Conclusion: Differential correlation analysis in combination with a graph-theoretical approach revealed a stronger small-world character in CPS subjects of the sensory and motor brain networks activated during the performance of a willed motor activity initiated by an emotional stimulus potentially as a result of compensatory mechanisms.