Early detection of breast cancer is essential to improve survival rates, and the use of Deep Learning techniques can revolutionize this field by offering more efficient and automated solutions. In this work, we propose a deep learning-based approach to breast cancer detection, using model architectures such as CNN, VGG16 and ResNet50. These models have been applied to the BreaKHis dataset, a histopathological dataset for the classification of breast lesions into benign and malignant. By integrating techniques such as transfer learning, we observed a significant improvement in model accuracy. Our results indicate that the use of deep learning can significantly improve the performance of breast cancer detection systems, with an accuracy of up to 87.43% for our CNN model. This study makes important contributions to the development of automated breast cancer detection systems, while paving the way for future research in medical imaging.