ABSTRACT
A number of approaches have been used in attempts to build useful computerized decision aids. This is especially true in the field of medicine where early techniques proved to be less than satisfactory. In this chapter, a system is described which utilizes a combination of rule-based methods encompassing approximate reasoning and neural network models. The method is illustrated in the development of treatment plans for patients with malignant melanoma. The rule-based approach allows expert input as well as provides explanation capabilities. Neural networks are used in two ways: first, as a device for extracting important decision making parameters from accumulated data, and secondly, to analyze laboratory tests directly to determine the likelihood of malignant disease. The resulting system has been tested and shows promising results for the establishment of treatment plans for this devastating disease.
