Medical expert system user interface

The theme for this issue of Artificial Intelligence in Medicine, for which I have graciously been selected as guest editor, is the user interface for medical expert systems. This problem has approximately the status of Gregor Samsa’s cockroach in Franz Kafka’s The Metamorphosis. The problem is too big and ugly to ignore, but not a subject for polite conversation in intellectual circles. By user interface, I am not speaking primarily of windows, colors, and other graphics, although these features have their place. As well-educated professionals, we come to our work with our own words and phrases in mind. If we enter these words into an expert system, then the system should be able to accept the words, even if they are in va.rious languages, follow various syntaxes, or draw vocabulary from various medical specialties. If our input is ambiguous, then the system may ask for clarification. Key is that the expert system input interaction begins with the user’s own words, not initially with a preset menu. In designing the expert system user interface, we may ask: In what language can we write our assertions and queries? English? German ? Japanese? SNOMED? What is the necessary syntax for these inputs? Struct,ured query la,nguage? PROLOG? Natural language? What medical semantics should already be known to the expert system? Anatomy? Pathology? Pharmacy? The past medical histories of previously admitted patients? Can we expect the expert system to render medical diagnoses ? If so, then who takes responsibility in the event of an unsatisfactory outcome ? If not, then can we at least expect that the expert system will alert us to a potential medical misadventure? Drs. Walters and Zhang address the user interface issue from the broadest perspective, namely the multilingual workstation. In this age of instant global communications, medicine is enriched by rapid exchanges of data and ideas across international boundaries. The authors discuss how linguistic differences can lead to differences in locating concepts and relationships in an expert system. Drs. Giere and Wakai propose a computer translator which translates short, declarative a,ssertions in medicine into PROLOG facts and rules. With this software, the physician can become a. logic programmer directly in natural language, rather than having to recall PROLOG’s cumbersome syntax rules. Dr. Sorace addresses the controversial area of expert system validation in blood bank donor deferrals. In the face of possible infectious disease contamination of donated blood (AIDS, hepatitis), there is a growing number and complexity of required tests, with corresponding chances for human error. This paper describes a rugged logic programming approach which can prevent programming and procedural catastrophes and correctly classify units of donated blood. Dr. Berman discusses on an important end-point for medical expert systems, namely the detection and prevention of apparent breeches in quality assurance. In this paper, quality assurance monitors are collected as part of the routine case reporting procedures in