Mapping the connectivity of the zona incerta with diffusion MRI: replicability and reproducibility

2186 

Abstract Submission 

Jason Kai¹, Roy Haast², Alaa Taha³, Violet Liu⁴, Ali Khan³, Jonathan Lau⁵

¹Robarts Research Institute, Western University, London, Ontario, ²Aix-Marseille University, Marseille, Provence, ³University of Western Ontario, London, Ontario, ⁴Western Univeristy, London, MT, ⁵Department of Clinical Neurological Sciences, Division of Neurosurgery, London, ON

Jason Kai  
Robarts Research Institute, Western University
London, Ontario

Roy Haast  
Aix-Marseille University
Marseille, Provence

Alaa Taha  
University of Western Ontario
London, Ontario

Violet Liu  
Western Univeristy
London, MT

Ali Khan  
University of Western Ontario
London, Ontario

Jonathan Lau  
Department of Clinical Neurological Sciences, Division of Neurosurgery
London, ON

The zona incerta (ZI) is a poorly understood deep brain region with growing evidence suggesting that it plays a crucial role across a wide range of brain functions [1], and is considered a candidate region for neuromodulatory therapies [2,3]. Advancements in MRI at ultra-high magnetic field strength (7 Tesla; 7T) have enabled direct visualization and differentiation of the human ZI and surrounding structures [4]. We previously demonstrated the feasibility of using diffusion MRI (dMRI) to map the structural connectivity (SC) of the ZI in vivo [5]. In this work, we investigate the replicability and reproducibility of identifying the internal organization of the ZI using 7T and 3T dMRI data from the Human Connectome Project (HCP).

SC between the ZI and the cortex was investigated using the minimally preprocessed HCP 7T (n=169), 3T (same subjects as 7T) and 3T test-retest (n=42) dMRI datasets. Cortical regions were defined in each subject's native space using volumetric HCP-MMP1.0 parcellations [6], while the ZI was previously derived from probabilistic parcellations [4]. Probabilistic tractography was performed using FSL's probtrackx [7] (default parameters unless otherwise indicated - 10000 samples per voxel, sampling in proportion along a set of fibre orientations, with distributions obtained with FSL's bedpostx [8]). Tractography was seeded from the ZI (thresholded at 50% and a 3 voxel radius dilation) to the target cortical regions, followed by transformation to the MNI152NLin6Asym template space. Spectral clustering (n=6 clusters) was performed by cosine-similarity based on connectivity patterns of the ZI to the cortex. Additionally, ZI connectivity gradients were extracted using BrainSpace [9]. Reliability was evaluated through comparisons between 7T and 3T (i.e. replicability), and test and retest 3T datasets (i.e. reproducibility). Centroid distances and Dice similarity from clustering-defined regions, as well as Procrustes disparity, comparing the connectivity patterns based on the first two gradients, were assessed.

Figure 1 exhibits the identified clusters in the (A) ZI and (B) their associated cortical regions. Evaluation results in C show good replicability (7T vs 3T), with an average centroid distance of 0.82 ± 0.47 mm and 1.00 ± 0.50 mm for left and right hemispheres, respectively. Similarly, good reproducibility (test vs retest) was shown with an average centroid distance of 0.37 ± 0.18 mm and 0.94 ± 0.46 mm. Dice similarity measures indicate good replicable overlap (0.78 ± 0.086 and 0.74 ± 0.14) and great reproducible overlap (0.83 ± 0.018 and 0.73 ± 0.079).

Figure 2 shows the two SC gradients explaining most of the variance (average eigenvalue of G1: 0.15 and G2: 0.08), in the (A) 3D MRI and (B) 2D gradient space. Procrustes analysis for left and right hemispheres, quantifying the similarity based on the 2D gradient distributions in C, revealed reproducible distributions among the 3T datasets. Notably in both analyses, the 7T dataset demonstrated the greatest difference relative to the various 3T datasets (Fig. 1C and Fig. 2D).

Reliable SC mapping of the ZI is crucial for furthering our understanding of ZI organization and to support its use for stereotactic neurosurgical planning. From both the cluster- and the gradient-based analyses, the largest differences were observed between the 7T dataset and the various 3T datasets. This can likely be attributed to dMRI acquisition as well as field strength related differences, and their impact on image and tractography quality [10]. Nonetheless, the high reproducibility among the 3T datasets supports the potential of SC-driven identification of the optimal ZI location for neuromodulatory therapies. Future analyses are required to validate these data-driven in vivo connections as well as to link these results to surgical outcomes.

Connectivity (eg. functional, effective, structural)

Subcortical Structures ²

White Matter Anatomy, Fiber Pathways and Connectivity ¹

Sub-Cortical

Other - Zona incerta, Structural connectivity, Cortical organization, diffusion MRI, Replicability, Reproducibility, Deep brain stimulation, neuromodulation, gradient