Algorithms of decision-making in intelligent advisory system for diagnostics of the mammary glands

Recently, for early detection of the temperature abnormalities in breast, that are indirect signs of tumors, specialists widely use the method of microwave radiometry. But the current diagnostic system developed on the basis of this method is designed for use by physician-mammologist with high qualifications, which narrows the scope of this method, and eliminates its unique features, especially those related to the identification of temperature anomalies in the early stages of the disease. To solve this problem we are developing intelligent advisory system, i.e. Expert System, that offers physicians a preliminary diagnosis and its rationale in a language understood by the physician.
Our work is the group project on the implementation of such a system. We consider the technology of the creation of two separate modules for decision-making subsystem of the designed system. These modules are based on a previously obtained by A.G. Losev and V.V. Levshinsky set of diagnostic features and implemented two algorithms for diagnosis of breast diseases. The first algorithm diagnoses using linear combinations of a set of signs, the coefficients of which are chosen to maximize the efficiency of the algorithm. The second algorithm is based on the use of neural networks.
As a part of the first module we established the diagnosis algorithm based on the parameters obtained by genetic algorithms and we adjust the optimal configuration of these settings to provide a good diagnostic result on the test sample.
When creating the second module we implemented several types of neural networks and conducted numerous computational experiments in order to select the optimal configuration of the neural network and its training algorithm.