Artificial Neural Networks in Biomedicine

Since the early 1990s, artificial neural networks play an increasing role in the development of new biomedical systems. In United States, over last decades, the granted biomedical patents that explicitly refer to artificial neural networks in their title, abstract or key references amount to about 50% of the total number of granted biomedical patents with a significant element of computational intelligence. Furthermore, the number of granted biomedical patents that explicitly involve artificial neural networks has quadrupled in recent four years. This observation greatly indicates the growing commercial interest in biomedical products involving artificial neural networks. As its title suggests, the edited book under review aims to provide a contour of the application of artificial neural networks in biomedicine. It covers a set of tutorial papers on established methods of best practice in artificial neural network design, and an extensive collection of case studies on commercially available products, recently-granted patents, and a wide range of applications. In detail, it contains 19 chapters arranged in 4 thematic sections. Each section includes a brief introduction to put the materials into context, and each chapter features at least one biomedical application of artificial neural networks. Roberts introduce the Bayesian framework for artificial neural networks, which is a generally accepted formalism for regularizing the multi-layer perceptrons to maximize the likelihood of finding the best network parameters while keeping the size of the network to a minimum. It is especially helpful that the authors use an example of diagnosing arm muscle tremor to illustrate the usefulness of the introduced technology in biomedical applications. In Chapter 2, W.G. Baxt, who is a pioneer in applying artificial neural networks to clinical medicine, highlights that artificial neural networks is a powerful nonlinear paradigm for the recognition of complex patterns. With a brief review on the application of artificial neural networks to the analysis of diversified human diseases, Baxt convincingly exhibits the robustness of artificial neural networks and its ability to often perform more accurately than the classic analytical approaches of the past in identifying disease and predicting its outcome. describe the process of using genetic algorithm to evolve the architecture of artificial neural networks so that the complexity of the model structure is minimized. Although this chapter is not as comprehensive as the one written by X. Yao [4], it is good in that the described process is illustrated by an example of predicting the occurrence of depression after …