Breast Nodule Classification: Single-modality Feature Mining and Multi-modality Information Fusion

Breast nodules are a significant early imaging manifestation of different breast diseases, especially malignant breast tumors. Accurate detection of them is important for the early diagnosis and intervention of these conditions. Presently, automatic detection techniques for breast nodules are classified according to the number of data types used in training into methods relying on single - modality data and those using multi-modality data. This paper gives an overview of the current status of single - modality and multi - modality methods in order to find safer and more efficient paths for the development of automatic breast nodule detection techniques. The analysis shows that methods based on single - modality data should focus on improving feature extraction, which the model does best. On the other hand, approaches using multi-modality data can make use of research from single - modality data to improve both the quality and quantity of models and fused data.