A Fuzzy, Non-Linear Similarity Measure for a Clinical Case-based Reasoning System

This paper presents a fuzzy, non-linear similarity measure designed for a clinical casebased reasoning system in radiotherapy treatment planning. The developed fuzzy similarity measure takes into account the distribution of attribute similarity values in the case base to ensure that the numerical values of the similarity between individual attributes are comparable and can be combined to give the aggregate similarity between two cases. Local fuzzy membership functions that are based on the attribute values of the target case are defined. The performance of the fuzzy similarity measure using local fuzzy membership functions is evaluated using real world brain cancer patient data. Preliminary experiments show promising results.