Breast cancer is a leading cause of morbidity and mortality among women globally, and pathology images are crucial for its diagnosis. In this paper, a breast cancer pathology image classification model based on ResNet-50 and VGG19 is constructed based on deep learning methods, and trained and tested using IDC breast cancer section images from the publicly available Kaggle dataset. A total of 13,403 preprocessed images were used in the experiment, and the performance was evaluated based on various classification and diagnostic indicators. The results show that ResNet-50 achieves 89.53% accuracy and 0.96 AUC value on the test set, and the overall classification performance is better than that of VGG19, especially in the recall rate and F1 score, while VGG19 is slightly higher in the precision rate. The model comparison verifies that deep convolutional networks have good potential for application in breast cancer pathology image recognition. Prospective studies may investigate the incorporation of magnetic resonance imaging alongside additional visualization data to construct an integrated multi-source framework, strengthening both the precision and practical value of computational approaches to breast cancer detection.