Domain Modelling for National Medical Practice Knowledge Banks (NMPKB)

The National Medical Practice Knowledge Banks (NMPKB) project, initiated in 1995 under the NIST-ATP Program for Information Infrastructure for Healthcare, has undertaken to address some of the large-scale needs of the medical community by developing a series of repositories of complex medical information in a variety of specialty areas. Each repository or "Knowledge Bank" will contain an integrated multimedia database of medical practice information as well as representative cases consisting of sample patient data, medical images such as MRI, CT or PET scans, surgical procedures, resulting diagnoses, treatment plans and outcomes. This information can take the form of text, speech, images, and video data. The technical goals of the project are to develop hardware and software tools to support the collection, organization, authoring and dissemination of this data. In addition, high-end software capabilities of the NMPKB include case-based retrieval of similar patient cases, synthetic interviews, training modules and integration of full-motion high-resolution video conferencing. These will allow each Knowledge Bank to be accessed for "anywhere, anytime" specialty consultations, expert patient management plans as well as realistic training and rehearsal capabilities.

The first Knowledge Bank under development is in the area of Neurosurgery, with a particular focus on brain tumors and stroke. The domain modeling task has been to identify the subset of the knowledge needed to support the functions of the Knowledge Bank and to determine a format for that knowledge. Examples of knowledge about medical practice include: vocabularies, concepts and relationships, facts, beliefs and opinions, guidelines for diagnosis and treatment as well as examples and anecdotes. In addition, knowledge about specific patient cases such as demographics, signs and symptoms, history, diagnoses, treatments and outcomes is needed as well as meta-knowledge used to annotate each piece of information with attributes such as sources, copyright, data type, level of belief, quality rating, medical peer review, intended audience and access privileges.