This is a picture of the opening screen of CaDDiS. The program uses standard Windows functionality for buttons, combo-boxes and lists. There are four data windows visible and three function buttons active in the bottom row of the window. These links will provide you with information about the functioning of each part of the system.

Advice button

The Advice button will generate the most likely diseases with added beliefs based on the Obvserved / Absent signs.

Reset button

The Reset button allows the user to remove all of the symptoms and enter a new query.

Help button

The Help facilities provide the user with a brief description of the system and how to use it.

Possible Signs window

The Possible Signs window lists a series of 27 clinical signs which might be observed in a sick cow. All signs may be accessed using the scroll bar to the right of the window. The Possible Signs window is active: Clicking on a clinical sign tells the system that this sign has been observed / absent in the cow whose disease is currently being evaluated. When a sign is entered in this way, it is removed from the Possible Signs window and transferred to the Observed / Absent Signs window.

Observed / Absent Signs window

The Observed / Absent Signs window serves two purposes: it records the clinical signs which have been selected by the user as being present or absent in the cow which is currently being examined, and, as an active window, allows the user to reject any or all of these previous entries. Clicking on a clinical sign in this window causes it to be removed from the current list. It will reappear in the Possible Signs window, allowing possible future re-entry.

Diagnoses / Beliefs window

The Diagnoses / Beliefs window serves to feed back the results which follow from the information which the user has entered into the program. The Diagnoses / Beliefs window lists those diseases which have a non-zero chance of being present in the animal given the clinical signs which have been observed. The belief rating of each listed disease is given in the Diagnoses / Beliefs window. In the initial window, 20 diseases will be visible. These diseases were selected, in consultation with veterinary experts, as the most serious diseases or syndromes affecting cattle in sub-Saharan Africa. As it stands, the system is not capable of evaluating the possibility of the cow having developed any disease other than the 20 diseases listed. Typically, as more clinical signs are entered into the Observed / Absent Signs window, the number of plausible diseases will fall, until a clear diagnosis can be made. The Diagnoses / Beliefs window provides a quantitative measurement of the level of belief estimated for each disease by the system, given the observed clinical signs. A simple explanation of how the program evaluates these 'probabilities' is given in the Dummy's Guide to Belief Networks. Note how the total of the beliefs is equal to one. This will always be the case; this fact immediately follows from our assumption that the cow is definitely afflicted with one of the twenty diseases under consideration. In the initial window each disease is given a belief rating of 0.05. This follows from the argument that, in the absence of any other information, we have no reason to favour the presence of any one disease option rather than another (a uniform prior in math-speak!).

When a new clinical sign is observed in the cow and this is added to the Observed / Absent Signs window, the Bayesian Belief network changes the posterior beliefs of the different diseases. The cumulative effect of the observed evidence is likely to be such that certain disease options may be excluded (belief = 0). For ease of interpretation, such diseases will be removed from the Diagnoses / Beliefs window and the Beliefs window (although they might reappear if some clinical signs were removed from the Observed / Absent Signs window.

When using the system, it is best to think of the beliefs as broad measurements of the relative likelihood of the cow having developed the different diseases. They are not probabilities in the normal sense of the term (if the cow has a particular disease, it has it. this is an event with probability equal to one: we just don't know which disease it is!) merely measurements of our uncertainty about which disease the cow may have.

In addition to these functions, the window also supports the operation of :

Clicking on a disease in the Diagnoses / Beliefs window opens a screen which will provide the user with additional information designed to help conclude a diagnosis.

A description of how to use CaDDiS is available, as are examples which illustrate how to interpret the findings of the system.