In order to make new and innovative rehabilitation programs for recovery of function after brain injury fully effective, we must provide access to all the critical factors that underlie the recovery process and the specific ways that therapeutic interventions modulate the recovery process across these multiple levels. We propose to develop tools that make available the synergistic combination of multiple sources and kinds of information, so that ultimately an individualized treatment plan can be generated based on experimental, theoretical, and clinical data, and current case information.
The various relevant information sources are distributed and organized in different manners, using diverse terminology. This causes syntactic, organizational, and semantic conflicts among various information sources. To support information unification and knowledge exchange among the heterogeneous information sources related to therapeutic interventions, we propose an approach for the semantic information management and integration of federated ontologies that characterize the information sources. This approach contributes to manage information using various perspectives on key concepts and relationships from neuroscience. The ultimate goals of this work are:
(1) to support the effective sharing of information across experimental and clinical settings, and (2) to provide a basis for the generation of a customized treatment plan for a patient, based on the combined knowledge gleaned from the various aspects of neuroscience research and clinical practice.

The research involves two essential aspects:
(a) construction of the databases and ontological descriptions (semantic meta-data) for various aspects of experimental, theoretical, clinical experience, and current case information; and (b) the federation of these varied data into a framework that facilitates their combined use.
In the research, we plan to work with an existing group of neuroscience experts, who are addressing various experimental, theoretical, and clinical approaches. This group is supported by an NIH investigational grant whose inter-disciplinary research Neuroplasticity and Stroke Rehabilitation – which we here view as heading under the broader banner of “Functional Neurotherapeutics” – has included not only research on molecular and cell biology, behavioral neuroscience, bioinformatics, computational modeling, virtual environment technology, haptics, biostatistics and physical therapy, but also development of a database for each phase of the research. The challenge here is to understand how further development of these databases can be coupled with research on their federation so that queries may be answered on the basis of data distributed across different resources. We focus on the ontology-based dynamic federation of multiple information sources and collections. Thus, our proposal involves essential computer science techniques and mechanisms, applied to the domain of functional neurotherapeutics.
The initial analysis categorizes the key information sources as follows. The first one is Clinical Database. It maintains clinical information to evaluate therapeutic interventions in individuals with physical disabilities. It includes imaging data, demographic data, clinical performance data, virtual reality data, and kinetics information data. The other is Animal Databases which is focused on rat experiments. This database is based on experiments on rats now, but it will be supposed to include monkey in the near future. Animal Databases involves demographic data, experimental data, assay data, and performance data. Another one is virtual reality experiment data for rehabilitation of stroke patients.
The database federation based on these disparate databases has two different challenges:
(i) the federation of animal and human data for use by basic researchers in such a way as to enrich the information available for treating patients; and (ii) the federation of general human data with data on a single patient to provide the clinician or therapist with information relevant to diagnosis and choice of therapy.
In particular, three incompatibilities among databases are detected:
(1) incompatibilities among brain regions from human and animals, (2) incompatibilities among human intervention for treatments and animal intervention for experiments, and (3) incompatibilities among assessments for human treatments and animal experiments.
The incompatibilities are inevitable because clinical database and animal database have different taxonomy and concepts. At first, human brain regions are classified differently from other animals’ brain regions. The other incompatibility is that there are no standard or classification among interventions from clinical database and animal database. The last one is incompatibilities among assessments exposed to the experiments.
Once database construction is accomplished, an ontology will be constructed for each database. These ontologies will be combined and integrated into a global ontology, which relates the information in the databases using semantic primitives relevant to neuroscience.
Ontology is constructed for each database. Ontology is effective means for capturing and representing real word knowledge in information system, contribute to unify multiple information representations and to keep the conceptual uniformity. Ontology can illustrate terminologies in a domain and their subsequent relations as well as the rule for combining terminologies and relations. Then, we will create a global ontology to integrate ontologies from each database. This integrative ontology has 2 mutually supportive aims: (a) to modify the ontology and thus the database schemas for the different databases to maximize their compatibility, while (b) developing tools to address remaining incompatibilities.
Determining the exact nature of these primitives is an essential part of this proposed research. For the process of ontology creation and integration, we propose to utilize a tool for managing dynamic ontologies and their inter-relationships that is being developed at USC – Ontronic. Once information is inter-related at the ontology/semantics level, web services are then employed to retrieve, combine, and present unified information from various sources to support clinicians and researchers in developing treatment plans, defining further studies, etc.