The first operator takes the input from the user and starts the PSM. The input is gathered by a dynamically expanding menu allowing textual summarization. It includes the patient history and presenting symptoms, vital signs, physical exam and the laboratory results. There is limited reasoning within the menus that enforces constraints among categorical values and appropriate precision for numeric values. The intention is to capture the information pertinent to the cardiovascular disease without requiring the system to do extensive reasoning beyond the program's intended medical domain and to provide physicians with a display of only the relevant patient information in an effective manner. Thus, it is assumed that the data has been interpreted and filtered by the user. For example, a single blood pressure is entered and assumed to be representative. Also, the program asks for interpreted findings on the electrocardiogram rather than the raw properties of the signal, as is more traditional. The physicians using this interface have found it to be an effective method of capturing and reviewing the data.
The evidence generator turns the data gathered from the menus into the initial PSM by using the data values to create value structures. These value structures cause a node structure to be created for each parameter state that could cause the value and those nodes in turn cause nodes to be generated for each possible cause of the node. Thus, by creating the value structures, the evidence generator creates the basic PSM structure used for most of the diagnostic reasoning.