It is proposed to develop a computerized expert system for analyzing and understanding the images of the nervous system generated by neuroimaging techniques such as computerized tomography (CT). A prototype version, called RAD (Radiologic Automatic Diagnosis), has been implemented. RAD will serve as an expert consultant in the field of neuroradiology, as an adjunct to computerized expert systems for medical diagnosis, and as an intelligent image processor supporting automated feature extraction from neuro-images. Modern neuroimaging techniques have revolutionized the practice of neurology and neurosurgery. The rapid proliferation of scanners, however, has outstripped the supply of experts with special training in scan interpretation. If a scanner could produce the differential diagnosis associated with the radiographs, it could aid radiologists in diagnostic tasks. Images received in digitized form from scanning devices will be segmented into objects after thresholding based on radiodensity (CT scans). This process will be aided by intelligent histogram optimizers and artifact removal programs. Analysis will be conducted in an object-oriented programming environment on objects represented in quadtree or octree format. RAD will employ clinical information, and knowledge bases in neuroanatomy, neuroradiology, neurology, and medicine to establish a context for analysis. This context will be instantiated in the form of assertions on a queue of beliefs in a truth maintenance system. Global and local image analyzers will also make assertions about neuro-images. The truth maintenance system will keep track of any contradictions which arise, and will provide forward chaining and backtracking explanation abilities of the reasoning leading to the differential diagnosis. At the present time, medical diagnostic expert systems, such as INTERNIST-I/CADUCEUS are dependent for its input upon the physician-user. The ability to directly interact with patient data in neuro-images would enrich the diagnostic foundation of the system and reduce obligate user interaction.

Agency
National Institute of Health (NIH)
Institute
National Library of Medicine (NLM)
Type
Research Project (R01)
Project #
1R01LM004431-01
Application #
3373688
Study Section
(SRC)
Project Start
1985-08-01
Project End
1988-07-31
Budget Start
1985-08-01
Budget End
1986-07-31
Support Year
1
Fiscal Year
1985
Total Cost
Indirect Cost
Name
University of Pittsburgh
Department
Type
Schools of Medicine
DUNS #
053785812
City
Pittsburgh
State
PA
Country
United States
Zip Code
15213
Banks, G; Short, P; Martinez, J et al. (1989) The alien hand syndrome. Clinical and postmortem findings. Arch Neurol 46:456-9