Associate Professor University of Maryland School of Medicine
Abstract Text: Introduction The individual thalamic nuclei play a central role as sensory and motor relay stations communicating between the sub-cortical regions, cerebellum, and cortex.1 In chronic traumatic brain injury patients, non-uniform atrophy has been seen throughout the thalamus with the nuclei bordering ventricles being most affected.2 Additionally, many treatments target specific nuclei to treat movement disorders and neuropathic pain, such as MR-guided Focused Ultrasound (MRgFUS) or DBS.3 However, sub-parcellation of the thalamic nuclei is challenging given the lack of contrast within the thalamus provided by conventional T1 or T2 images. In this study, we propose using synthetic-TI images from the T1w MPRAGE and FGATIR sequences to accurately visualize and delineate thalamic nuclei assisted by the Morel stereotactic atlas.4 We further report multi-parametric tissue properties of thirteen thalamic nuclei based on T1, local magnetic susceptibility, and water diffusion measures for control subjects and mil TBI patients.
Methods Imaging: MRIs were acquired on a Siemens 3T PrismaFIT scanner using a 64-channel head and neck coil. 3D MPRAGE images were acquired with two TI’s, 400ms and 1400ms, providing white matter-nulled (FGATIR) and T1-weighted (T1-MPRAGE) contrast respectively. QSM images were acquired with a 3D flow-compensated gradient echo sequence. DKI images were acquired using an SMS EPI sequence. Subjects: MRI data were collected on: (1) eleven orthopedic-injury control subjects (3 females, average age: 41.6±13.6 years, range: 21-59 years) from the University of Maryland Medical Center with no history of neurological issues or head injuries; and (2) eleven mild TBI subjects (5 females, average age: 37.8±14.3, range: 21-59 years) with a Glasgow Coma Scale (GCS) between 13-15 upon admission to UMMC and documented loss of consciousness, positive head CT, or facial trauma. Data analysis: Synthetic TI images were produced by co-registering the FGATIR and T1-MPRAGE images. All images were first aligned to the axial AC-PC aligned T1w images, which were further warped to the MNI152 standard atlas. QSM maps and DKI maps were reconstructed offline and warped to the MNI space. Susceptibility values were further separated into paramagnetic and diamagnetic source values. The digitized Morel atlas7 in the 1mm MNI152 atlas space was used as the initial thalamus mask. Neighboring smaller nuclei similar in function and T1 value were combined into composite nuclei. The T1 map and the SynTI images were primarily used to manually adjust each nucleus mask.
Results In the control group, the average standard deviation (SD) and coefficient of variation (CV) for T1 values for all nuclei were lower for the corrected labels (SD: 29.34-105.10ms, CV: 3.45-10.51%) compared to the uncorrected Morel atlas (SD: 31.15-242.34ms, CV: 3.66-21.36%). For both control and TBI groups, the ventral nuclei (VA, VLP, VPL, VPM) have lower T1 values and higher diamagnetic source values compared to other nuclei. The ventral nuclei demonstrated high FA and MK values due to the high density of somatosensory and motor axons.5,6 The AN and lateral dorsal (LD) exhibited high ADC values because of their proximity to the ventricles. In this small group of patient data, we did not observe a difference in T1, QSM, or DKI values between the control and TBI groups. However, the TBI group displayed greater variance in QSM value compared to the control subjects.
Discussion Our results suggest that synthetic MPRAGE images can be helpful in delineating thalamic nuclei as outlined in the stereotactic Morel atlas. Distinct tissue properties reflected with T1 and DKI values were revealed in each nucleus, allowing for the development of automatic feature-based thalamic nuclei segmentation. With our current sample, no significant differences between mild traumatic brain injury patients and non-head injury control subjects were found. Active recruitment of normal control subjects and TBI subjects is ongoing to confirm the validity of the delineation method and to further investigate possible changes in the thalamus in the mTBI population. The manual thalamic labels generated through this process will also serve as ground truth to train automatic thalamic segmentation algorithms in the future.