International Science and Technology Journal

Breast cancer is a prevalent and complex disease that necessitates accurate classification for timely diagnosis and effective treatment. This paper proposes a breast cancer classification system that leverages the ResNet-50 model and incorporates transfer learning. The system utilizes two separate ResNet-50 models: one for normal and abnormal classification and another for benign and malignant classification within the abnormal category. The experimental evaluation employs three datasets: MIAS, DDSM, and CBIS-DDSM. The wavelet transform technique is applied as a pre-processing step to enhance mammography images. The system achieves outstanding performance with high accuracies of 99.48% and 99.25% for the first and second models, respectively. Additionally, sensitivity and specificity values of up to 99.32% and 99.17% for the first model and 99.07% and 98.89% for the second model are obtained. Notably, the DDSM dataset consistently outperforms other datasets in both models. The proposed system demonstrates the efficacy of the ResNet-50 model and the wavelet transform technique in breast cancer classification, offering a promising tool for early detection and precise diagnosis. KEYWORDS: Breast cancer • ResNet-50 model • transfer learning • Daubechies wavelet transform • mammography • wavelet transform.