The application of machine learning techniques to problems in medicine are now becoming widespread, but the rational and advantages of using a particular approach is not always clear or justified. This chapter describes the application of a version of Cartesian Genetic Programming (CGP), termed Implicit Context Representation CGP, to two very different medical applications: diagnosis and monitoring of Parkinson’s disease, and the differential diagnosis of thyroid cancer. Importantly, the use of CGP brings two major benefits: one is the generation of high performing classifiers, and the second, an understanding of how the patient measurements are used to form these classifiers. The latter is typically difficult to achieve using alternative machine learning methods and also provides a unique understanding of the underlying clinical conditions.
|Title of host publication
|Inspired by Nature
|Susan Stepney, Andrew Adamatzky
|Number of pages
|Published - 2018
|Emergence, Complexity and Computation