Breast cancer is one of the most common an‎d deadly diseases among women, where early detection plays a crucial role in reducing mortality rates. Mammography images are recognized as an effective tool for the early diagnosis of breast cancer; however, accurate analysis an‎d identification of patterns in these images require advanced image processing an‎d machine learning techniques. This study aims to improve the accuracy an‎d efficiency of breast cancer detection systems by extracting important features from mammography images using wavelet transform an‎d applying various machine learning an‎d deep learning models. For this purpose, two well-known datasets, MIAS an‎d DDSM, were used for model training an‎d eva‎luation. Initially, key features were extracted from the mammography images using wavelet transform. Then, three main approaches were eva‎luated for breast cancer detection: the maximum learning machine method, traditional machine learning methods (including Ran‎dom Forest, Support Vector Machine, an‎d K-Nearest Neighbors), an‎d hybrid models combining machine learning with deep neural networks. All these methods were applied after feature extraction using wavelet theory. The results, eva‎luated using various performance metrics such as accuracy an‎d precision, demonstrate that the hybrid model combining Support Vector Machine an‎d deep learning achieved the best performance. This model showed high accuracy, particularly in complex an‎d rare cases, reaching an overall accuracy of 90%. The hybrid approach not only improved the detection of positive cases (malignant tumors) but also showed better generalizability to new data. This method can be considered an effective tool for the automated diagnosis of breast cancer from mammography images in clinical an‎d hospital settings, ultimately helping to reduce human error an‎d improve diagnostic outcomes. Keywords: Breast cancer, Mammography images, Wavelet transform, Machine learning, Deep learning

مريم ذكري

الهام محمودزاده

فريد شيخ الاسلام , جعفر قيصري