A Structural Connectivity Approach to Identify Human Paraventricular Thalamic Nucleus [Presidential Award Proceedings]

<p>Introduction</p><p>　Accumulating evidence has connected the dysfunction of the paraventricular thalamic nucleus (PVT) with psychiatric disorders, including anxiety and major depressive and bipolar disorders. This study aimed to identify PVT and its specific connectivity patterns using diffusion MRI tractography and a k-means clustering algorithm.</p><p>Methods</p><p>　The diffusion MRI data of 100 healthy individuals (55 women, mean age, 28.5±3.9 years) from the Human Connectome Project and 10 healthy individuals (5 women, mean age, 44.6±13.9 years) from the Juntendo University were assessed as the first and test cohorts, respectively. Probabilistic tractography was performed based on the Bayesian Estimation of Diffusion Parameters using FreeSurfer's magnocellular subdivision of the mediodorsal thalamus (MDm), which is predicted to contain PVT, as seed regions, and cortical and subcortical areas as targets. The MDm was subsequently segmented into two regions using a k-means clustering algorithm based on the frequency matrix of fibers originating from each seed voxel in the MDm and reaching the specific target. Furthermore, the partial Pearson's correlation tests were used to evaluate the associations of the number of streamlines in the structure considered to be PVT with alcohol-related scores and adjusted for age, sex, and intracranial volume.</p><p>Results</p><p>　In both the cohorts, we were able to segment the MDm into dorsomedial and ventrolateral regions consistently (Fig. 1A). The tracts originating from the dorsomedial part of the MDm predominantly projected to the limbic areas, including the amygdala, nucleus accumbens, and anterior cingulate cortex. Thus, they most likely represent the PVT (Fig. 1B). Significant correlations were found between the number of streamlines in the ventromedial part and alcohol−related scores (Fig. 2).</p><p>Conclusion</p><p>　Structural connectivity features from probabilistic tractography and k-means clustering enabled segmentation of PVT.</p>