Automated MRI Segmentation of Brainstem Nuclei Critical to Consciousness

Area of Science:

• Neuroimaging
• Neuroscience
• Brainstem Anatomy

Background:

• Mapping cortical networks for consciousness is advanced, but subcortical arousal networks remain understudied.
• Lack of robust segmentation for ascending arousal network (AAN) nuclei limits research into wakefulness and consciousness disorders.
• Automated segmentation of AAN nuclei is vital for understanding consciousness physiology and pathophysiology.

Purpose of the Study:

• To create a probabilistic atlas and Bayesian segmentation algorithm for brainstem AAN nuclei.
• To enable accurate and automated identification of AAN nuclei in neuroimaging data.
• To facilitate the study of human consciousness and its disorders.

Main Methods:

• Created a probabilistic atlas of 10 AAN nuclei using diffusion MRI from ex vivo human brain specimens.
• Manually annotated neuroanatomic boundaries using high-resolution MRI and immunostains.
• Developed a Bayesian segmentation algorithm using the atlas as a generative model for adaptive nucleus identification.

Main Results:

• The segmentation method demonstrated high accuracy on in vivo MRI scans of healthy individuals and TBI patients.
• Achieved high test-retest reliability across different MRI contrasts (T1 and T2).
• The algorithm detected volumetric changes in Alzheimer's disease and magnetic susceptibility differences in traumatic coma patients.

Conclusions:

• The developed probabilistic atlas and Bayesian segmentation tool advance the study of human consciousness.
• This method enables the detection of AAN nuclei changes in neurological disorders.
• The tools are publicly released to foster further research in consciousness and its disorders.