Evolutionary Fuzzy Extreme Learning Machine for Mammographic Risk Analysis

Mammographic risk analysis is an important and challenging issue in modern medical science; research and development in this area has recently attracted much attention. Many efforts have been devoted to achieving a higher accuracy in such analysis. This paper presents a novel approach for automated analysis of mammographic risk, in support of human consultant estimation of such risk. The underlying approach is general, it combines evolutionary computation with extreme learning machine to efficiently train effective fuzzy systems. The proposed approach is experimentally compared to a number of state-of-the-art learning classifiers that can also be adopted to analyze mammographic risk. The significance of this work is highlighted by its improved performance over the alternative approaches, measured using criteria such as classification accuracy and confusion matrices. The results demonstrate that for the problem of mammographic risk analysis, evolutionary fuzzy extreme learning machine entails such performance both at the overall image level and at the level of individual risk types.