Intelligent Controller for Clinical Science Experiments

We will perform proof-of-concept research on a Real-Time Expert System that will entail several parts: 1) capture data from the computer that presently controls a clinical laboratory's testing or research ; 2) perform sophisticated analyses of incoming results; 3) make routine test or experiment control decisions of the sort presently made by the clinician-scientist, basing them on the clinician's priorities, reasoning methods, statistical criteria and rules of thumb, all of which are encoded in the expert system's knowledge base as a hierarchical set of interdependent hypotheses; 4) either directly command the present control computer to implement its decisions, allowing the clinician or technician to manipulate priorities and high level control variables to influence the testing or research protocol; or 5) convey recommendations to the clinician or technician, so she/he may make the ultimate decisions. We expect that our intelligent controller will autonomously determine when statistically valid results have been accumulated, recognize and repeat defective or failed measurements, appropriately alter experiment or test parameters and protocol depending on accumulated results, and in the case of clinical research, recognize unexpected opportunities and adjust protocol to take advantage of them. Our aims in Phase I are to demonstrate feasibility of our product by modifying the knowledge base of our prototype intelligent controller to receive data over a network connection to a laboratory computer at Mass Eye and Ear Infirmary and render experiment-control recommendations concerning a particular auditory neurophysiology experiment. We will evaluate our distributed-computing control concept, evaluate the quality and timeliness of experiment control recommendations, identify modifications of architecture to facilitate our Phase II research, and identify software tools that we must construct to facilitate our Phase III efforts. Our system should improve experiment or test flow decisions by increasing the sophistication of analysis on which they are based; it should improve productivity and cost-effectiveness of clinical research that is amenable to computer control.