Artificial neural networks in urologic oncology
An artificial neural network (ANN) is a complex computer system that is designed to replicate human decision making by modelling the human neuron [1-3]. The human brain is made of neurons and millions of multiple synapses, and it is believed that with learning, weighting of individual synapses affects the final decisions we make. As input neuronal information is passed through on multiple simultaneous pathways to converge on the final output neuron, it is under the influence of either inhibitory or stimulatory input at each synapse. For a computer to model this process, a neural network is created as a network of many very simple processors, each with its own local memory, linked by unidirectional connections that carry numeric data. The units operate only on their own local data and on the inputs they receive via the connections. Each node receives input from other nodes and, with learning, it changes the weights of the incoming information to produce an output that most closely represents the known outcome. When the sum of the weights exceeds a predetermined threshold, the node fires; otherwise it remains quiet. The combination of the firings of each neuron determines the final decision of the output neuron. Within this system are often hidden layers of neurons that further influence the final outcome. This is unlike conventional computers, where there is a central processor with central memory that processes information sequentially using a defined set of rules. Figure 8.1 represents a model of a neural network [4].