PROJECT SUMMARY/ABSTRACT
The thalamus is associated with critical neurological functions like regulation of consciousness, sleep, arousal,
and alertness in addition to relaying signals to the cortex. It is divided into multiple functionally specialized units
called thalamic nuclei which have been implicated in several psychiatric and neurodegenerative diseases such
as essential tremor, Parkinson’s disease, and schizophrenia.
Automated segmentation of thalamic nuclei from MRI data is not commonplace, due to their poor visibility in
conventional MRI. Segmentation methods based on Diffusion Tensor Imaging (DTI) have been limited by the
spatial resolution of the echo-planar imaging (EPI) acquisition and poor diffusion anisotropy in the grey-matter
dominated thalamus. As a result, most neuroimaging studies treat the thalamus as a single entity,
characterizing whole volume changes and using the whole thalamus as a seed for connectivity analyses,
significantly reducing sensitivity to nuclear-specific changes in pathology. We have developed automated multi-
atlas as well as deep-learning based thalamic nuclei segmentation techniques based on a novel white-matter-
nulled contrast scheme. The purpose of this grant is to develop the next-generation methods for thalamic
visualization and segmentation using multi-contrast imaging and cutting-edge image processing techniques
and testing it on pediatric as well as geriatric populations. This will be achieved using the following aims:
a) Development of a novel fast motion-robust multi-contrast imaging sequence which will provide co-registered
susceptibility weighted and MPRAGE images with different contrasts (e.g. white-matter and CSF-nulled).
b) Acquisition and creation of age-stratified atlases using the proposed multi-contrast sequence
c) Development of a multi-contrast deep-learning based automatic segmentation scheme
d) Development of a contrast-synthesis strategy to segment conventional MPRAGE and SWI data but
leveraging the multi-contrast atlas developed
e) Documenting changes in anatomical, functional and structural connectivity using data from publicly available
OASIS (geriatric) and ABIDE (pediatric) databases using the proposed segmentation methods.
The segmentation methods developed here can be used characterize thalamic atrophy in normal aging and in
disease populations with high sensitivity. The project is expected to yield new MR imaging biomarkers which
could be used in future studies for the identification and evaluation of novel therapeutic targets.