Breast Cancer diagnosis based on a lightweight convolution network

Breast cancer remains one of the leading causes of mortality among women worldwide, underscoring the urgent need for accurate and dependable diagnostic systems. Early detection significantly improves treatment outcomes, making the development of effective computer-aided diagnostic tools essential. In this study, we propose a hybrid optimization-enhanced Convolutional Neural Network (CNN) for breast cancer classification, where a lightweight CNN architecture is integrated with the Sparrow Search Optimization Algorithm (SSOA). The role of SSOA is explicitly emphasized, as it optimizes key hyperparameters and strengthens the model’s generalization capability. To increase dataset variability and further improve robustness, advanced data augmentation techniques are incorporated, while a rigorous 10-fold cross-validation procedure is employed to ensure performance stability and reduce overfitting.Experimental findings show that the base CNN model achieves an accuracy of 97.3%, which increases to 98.5% when evaluated using 10-fold cross-validation. These results outperform several widely used deep learning architectures, including ResNet, VGG16, and GoogLeNet, demonstrating the effectiveness of the proposed hybrid SSOA-enhanced approach. Overall, the revised Abstract now clearly communicates the novelty and significance of combining a lightweight CNN with SSOA to achieve more accurate and reliable breast cancer detection.