Method and apparatus for training and operating a neural network for detecting breast cancer

A digital storage medium containing a program that, when executed by a general purpose computer, causes the computer to perform a method for using a neural network containing a mixture of experts to predict breast cancer from skin surface potentials comprising the steps of:

• recalling training data containing skin surface potentials, training patient information containing menstrual cycle information for each patient, and a known correct diagnosis for the training data;
• using the training data, training patient information and the known diagnosis to train a gating network and a plurality of expert networks;
• applying new data containing skin surface potentials to the expert networks;
• applying new patient information corresponding to the new data to the gating network;
• and
• determining a probability that the new data represents malignant breast cancer using trained gating network and expert networks.

A neural network containing a mixture of experts to predict breast cancer from skin surface potentials comprising:

• a memory for storing training data containing skin surface potentials, training patient information containing menstrual cycle information for each patient, and a known correct diagnoses for the training data;
• a processor for recalling the stored training data, training information and diagnoses and using the training data, training patient information and the known diagnoses to train a gating network and a plurality of expert networks;
• means for applying new data containing skin surface potentials to the expert networks and for applying new patient information corresponding to the new data to the gating network;
• and
• said processor determines the probability that the new data represents malignant breast cancer using the trained gating network and expert networks.