DE-Ada*: A novel model for breast mass classification using cross-modal pathological semantic mining and organic integration of multi-feature fusions

•We propose an effective model called the DE-Ada* for breast mass classification.•We use cross-modal pathological semantic to alleviate the overfitting problem.•We make full use of the implicit complementarity among diverse image features.•We propose a novel Kiviat diagram-based metric to evaluate each model.•We propose an end-to-end breast mass classification system based on the DE-Ada*.