Clinical Computational Anatomy Core
Introduction
The challenge of understanding the human brain, arguably the most complex and least understood of all body organs, requires the shared knowledge and expertise of the entire spectrum of medical researchers and physicians; it is impossible for any one individual to comprehend and assimilate more than a small fraction of the available neuroscientific information. One powerful approach to meeting NIH’s goal of “sharing ideas and resources to promote discovery” is the development of computer-based knowledge systems capable of organizing and linking together morphological and functional information about the brain. Such systems can present a unified view of knowledge about structures and behavior ranging from molecules, to cells, to neural circuits, to anatomo-functional systems, to basic neurologic function, to human behavior. Using such a system, a wide range of medical professionals can both contribute to and benefit from a pool of common knowledge, amplifying the value of federal investment made in individual research efforts. A computer repository of brain knowledge is most useful and expressive when it not just includes, but also bridges, the morphologic and functional aspects of neural systems. Indeed, function can only be understood in terms of the physical structure that underlies it. Conversely, since the role of morphology is to support function, its significance cannot be fully appreciated, unless it is approached from a functional perspective. Under the direction of Core PI Ion-Florin Talos, the CCA Core is working to develop an integrated anatomic atlas that includes elements of each of these areas:
- a volumetric brain atlas based on high resolution, multimodality MRI scans of a living, normal subject
- expert labeling of the anatomic structures imaged in these scans
- linkage of these labels to established, controlled vocabularies and ontologies of anatomy
- development of a new functional ontology describing the human motor system and the effect of neurodegenerative diseases such as Parkinson's Disease.
Aim 1 - Develop an ontology-based symbolic model describing the basic functional organization of the motor system
Parallel with the unprecedented evolution of neuroimaging techniques, neuroscientific data has accumulated at an exponential rate, making it impossible for any one individual to comprehend and assimilate more than a fraction of the available information. Computer-based systems for organizing and disseminating medical knowledge have emerged in response to this challenge. Ontologies are a methodology for organizing this information glut into coherent, explicit knowledge repositories. The term ontology is defined as an explicit, structured description of a knowledge domain. Computational ontology systems are a rapidly advancing mechanism for expressing the common vocabulary, shared understanding, and relationship of different concepts in a field in a way that is useful for both human understanding and automated computer reasoning. The combination of established neuroanatomic knowledge with an ontological representation provides a better understanding of the physical organization of the nervous system, a mechanism for integrating morphologic and functional aspects of the brain, and a framework for organizing many forms of neuroscientific data in an extensible way that can be used as a resource for dissemination, decision-making, and further research.
Our modeling framework is rooted in the organizational principles of functional connectionism (Wernicke, Sherrington, Cajal) and revolves around the central concept of neural network – i.e.. a group of interconnected neurons performing elementary signal processing operations as part of a specific brain function. It is based on a disciplined modeling approach, relying on a set of declared principles, a high-level schema, Aristotelian definitions and a frame-based ontology authoring system (Protégé 2000). It describes both meronymic (“is part of”) and functional (physiologic) relationships – i.e. physiologic effects, such as excitation and inhibition, exerted by “origin” neural network nodes on their “target” network nodes, conveyed via neural network connections.
Aim 2 - Create an ontology-enhanced, three-dimensional brain atlas and link the symbolic model of motor neural networks with the 3D-brain atlas
The goal of creating an ontology-augmented brain atlas will be accomplished in three steps: creating a detailed, three-dimensional brain atlas from high-resolution anatomic and Diffusion Tensor MRI, linking the atlas with the functional ontology, and modifying the visualization software (3D-Slicer) to provide a user-friendly interface for the ontology-augmented brain atlas.
The MRI data that serves as basis for the construction of the multi-modal atlas was acquired in a healthy male volunteer, during several two-hour scanning sessions, at the Martinos Center for Biomedical Imaging. A multi-channel head coil and a 3T Siemens scanner were employed. The data consists of multiple sequences: volumetric T1-weighted (multiple whole-head MPRAGE acquisitions, voxel size .75 mm), volumetric T2-weighted (whole head T2-FSE, voxel size .75 mm), and DTI (multiple whole brain acquisitions, employing 120 diffusion sensitizing gradients; voxel size 1mm, no gap).
Aim 3 - Validation of the integrated symbolic-geometric model
According to the current pathophysiologic models of movement disorders, the net excitation in the internal pallidal segment directly influences the net excitation of the motor cortex, via the ventral anterior thalamic nucleus. The internal pallidal segment exerts an inhibitory effect on the motor cortex. In hypokinetic disorders, such as Parkinson’s disease, the net excitation in the internal pallidal segment is increased, whereas it is decreased in hyperkinetic disorders, such as Hemiballism and Chorea. For instance, in Parkinson’s disease, the increased excitation level in the internal pallidal segment is due to decreased activity in the direct pathway, resulting from degeneration of dopaminergic neurons in the substantia nigra pars compacta. In Hemiballism, there is a decreased level of excitatory influence from the subthalamic nucleus on the internal pallidal segment, resulting in an increased level of activity in the motor cortex. In order to reproduce these relationships between the nuclei (nodes) of the motor initiation network under normal conditions, as well as in movement disorders, each excitatory connection in the motor initiation network can be attributed a positive value, and each inhibitory connection in the same network can be attributed an arbitrary negative value. An automated reasoner then computes the net excitation level in the motor cortex. If it differs from the normal value, the reasoner identifies the site of increased excitation/inhibition in the basal ganglia neural network, responsible for the increase/decrease of net excitation in the motor cortex. Once identified, the responsible structure is highlighted in the brain atlas as a possible target for therapeutic intervention. Additionally, the system will identify alternative therapeutic targets, i.e. basal ganglia network nodes whose surgical inactivation would lead to a normalization of the net excitation in the motor cortex.
Bibliography
- Talos IF, Rubin DL, Halle M, Musen MA, Kikinis R. A prototype symbolic model of canonical functional neuroanatomy of the motor system. J Biomed Inform. 2008 Apr;41(2):251-263. PMID: 18164666. PMCID: PMC2376098.
- Rubin DL, Talos IF, Halle M, Musen MA, Kikinis R. Computational neuroanatomy: ontology-based representation of neural components and connectivity. BMC Bioinformatics 2009; 10(Suppl 2):S3. PMID: 19208191. PMCID: PMC2646240.