Classification of potential multiple sclerosis lesions through automatic knowledge extraction by means of differential evolution

In this paper a classifier, designed by taking into account the user-friendliness issue, is described and is used to tackle the problem of classification of potential lesions in Multiple Sclerosis. This tool is based on the idea of making use of Differential Evolution (DE) to extract explicit knowledge from a database under the form of a set of IF-THEN rules, can use this set of rules to carry out the classification task, and can also provide clinicians with this knowledge, thus explaining the motivation for each of the proposed diagnoses. Each DE individual codes for a set of rules. The tool is compared over a database of Multiple Sclerosis potential lesions against a set of nine classification tools widely used in literature. Furthermore, the usefulness and the meaningfulness of the extracted knowledge have been assessed by comparing it against that provided by Multiple Sclerosis experts. No great differences have turned out to exist between these two forms of knowledge.