Breast cancer is still a major worldwide health concern, so developing precise and effective diagnostic methods is essential. Although they are essential for early diagnosis, traditional medical imaging methods like mammography have drawbacks include human error, inconsistent results, and trouble spotting small abnormalities. Additionally, noise and poor image quality are still major problems in breast cancer image classification, which can impair automated diagnostic systems’ performance. This study presents a hybrid approach for breast cancer classification by combining the EfficientNetV2S deep learning model for feature extraction and classification with Contrast Limited Adaptive Histogram Equalization (CLAHE) for image enhancement. CLAHE is used to improve feature extraction, lessen the effect of noise, and make tiny details in mammography pictures more visible. For binary classification (benign vs. malignant), the EfficientNetV2S model is used, which is renowned for its strong performance and computational efficiency. According to experimental findings, adding CLAHE to the preprocessing pipeline greatly improves image quality, which raises classification robustness and accuracy. This approach leverages advanced image preprocessing and deep learning to address limitations of traditional breast cancer detection methods. The findings indicate the potential for this solution to provide reliable and precise breast cancer detection and diagnosis in clinical applications.