<p>Ultrasound examination is a common method for detecting breast tumor. Most existing computer-aided diagnosis (CAD) methods for tumor detection and lesion classification rely on the extraction of features from the spatial domain. However, the traditional spatial domain methods is hard to detect subtle structural or early pathological changes in tissues, and intensity variations in the spatial domain are insufficient to accurately reflect the microstructure and heterogeneity within tissues. In this paper, a dual-frequency feature enhancement network (DFENet) for breast tumor classification is proposed to enhance the remarkable feature of tumor boundaries and minimize the impact of high-frequency noise. To obtain robust feature representation, the dual-frequency feature fusion (DFF) module is proposed to enhance efficient representation of the edge and texture features of breast tumor using. To further enhance the model’s multi-scale feature extraction capability, the multi-scale global–local fusion (MGF) module is proposed, which merges global and local information of different scales, enhancing the comprehension and classification capabilities regarding lesion characteristics. The synergy of the DFF module and the MGF module significantly improves the network’s ability to capture complex lesion images’ edges and textures, further refining the detection and differentiation of breast tumor. Experimental results indicate that the proposed network obtains the best performance among the state of the art in the datasets of BUSI and GDPH&amp;SYSUCC. The code is shared in <a href="https://github.com/WTU-MIS-Laboratory/DFENet">https://github.com/WTU-MIS-Laboratory/DFENet</a>.</p>