Neural network modeling with medical data

Neural networks have been shown to be useful in the design of medical decision-making aids. These models are easily adaptable to new applications. The use of neural networks for medical decision-making making has been limited in applications in which most data items are descriptive rather than numeric. This work focuses on pre-processing of medical data in order to properly define input parameters for the neural network model, including handling unordered data and complex data structures. In addition, methods for dimensionality reduction are needed. The refined parameters are used as input for the Hypernet learning algorithm developed by the authors. The method is illustrated in a model for diagnosis of dementia.