Machine Learning-Based Diagnostic Model for Accurate Prediction of Breast Cancer Using Immunohistochemical Images

Background: Breast disease, particularly breast cancer, ranks among the most prevalent malignancies affecting women globally. Accurate clinicopathological diagnosis is critical for early detection and prognostication of breast cancer. This study aimed to establish an ultrasensitive diagnostic model utilizing machine learning to assist in breast cancer pathology. Methods: By integrating bioinformatics, we identified four targets-DPP3, KIF4A, TK1, and UBE2C-with significantly higher expression levels in breast cancer tissues compared to adjacent normal tissues, supported by corresponding immunohistochemical staining images obtained from the HPA database. Using machine learning, we developed a pathological image recognition algorithm for breast cancer. Results: Our findings revealed that the diagnostic accuracy for DPP3 and KIF4A was significantly superior, achieving 93% and 92%, respectively, while TK1 and UBE2C attained accuracies of only 76% and 62%. However, the combined diagnostic efficacy of TK1 and UBE2C increased to 99%. Conclusion: This study highlights the potential of machine learning algorithms in the classification and diagnosis of breast cancer pathology images, emphasizing the importance of integrating bioinformatics with machine learning to enhance early diagnosis and facilitate personalized treatment strategies for breast cancer.