Breast Cancer Microcalcification Classification via a Hybrid Deep Learning Model with Discrete Wavelet Transform and Attention-Enhanced ResNet-50

Abstract

Breast cancer remains a major global public health concern with steadily increasing incidence rates. Early and accurate detection plays a crucial role in improving treatment outcomes and patient survival. Among diagnostic challenges, the reliable classification of breast calcifications, particularly microcalcifications, remains difficult due to tumor heterogeneity and the complexity of mammographic imaging. To address this issue, we propose a novel hybrid deep learning framework for the classification of breast microcalcifications in mammography. The proposed model integrates three key components: Discrete Wavelet Transform (DWT) for multi-resolution feature extraction, Squeeze-and-Excitation (SE) attention mechanisms for adaptive feature recalibration, and a ResNet-50 backbone for deep feature representation. Prior to classification, mammographic images undergo preprocessing, including noise suppression and contrast enhancement. Experimental validation was conducted on the publicly available CBIS-DDSM dataset. The proposed DWT and SE-augmented ResNet-50 model achieved a maximum classification accuracy of 95.74%, outperforming the baseline ResNet-50, which attained 93.79%. These findings demonstrate that the proposed hybrid approach enhances the diagnostic performance of mammographic image analysis and provides a promising tool for supporting clinical decision-making in breast cancer detection.