This proposal centers on investigating the design and feasibility of various user interfaces for the CONTRAST expert system currently under development at the Pittsburgh NMR (Nuclear Magnetic Resonance) Institute. CONTRAST is a knowledge based computer consultant for assisting NMR radiologists in the tast of selecting scanning parameters which optimize diagnostically useful contrast in the resultant images. The Phase I study proposed herein will explore feasible alternatives for user interface designs to this NMR expert system. Specifically, we will consider the graphical display of information, and its relation to internal knowledge representations the use of natural language processing computer understanding of user goals, plans and intentions for dialogue management. Phase II will undertake a full integrated pilot implementation of the Phase I design. The implemented user interface will be refined and tested with the CONTRAST expert system, and used by radiologists as a tool to assist them in NMR imaging. Successful completion of this research in the Phase II study would extend the current knowledge of the use of discourse structure in natural language interface design. Further, it would lead to a commercial software product useful to clinical radiologists.
