Choquet Integral based Feature Selection for Early Breast Cancer Diagnosis from MRIs

Abstract

International audienceThis paper focuses on breast cancer of the mammary gland. Both basic segmentation steps and usual features are recalled. Then textural and morphological information are combined to improve the overall performance of breast MRI in a computer-aided system. A model of selection based on Choquet integral is provided. Such model is suitable when handling with a weak amount of data even ambiguous in some extent. Achieved results compared to well-known classification methods show the interest of our approach

Choquet integral with stochastic entries

“…Since almost 3 decades, the Choquet integral has become a major "actor" in the field of aggregation function and decision making [6,7]. Thanks to theoretical works, many applications have been developed in finance [3], sustainable development [9,1], image processing [12], recommender system [4], clustering [10], deep-learning [2], to only cite few in recent years. Despite the amount of studies, the question of uncertainty in the Choquet integral remains under-considered.…”