Artificial intelligence in the diagnosis of breast pathologies: A literature review
DOI:
https://doi.org/10.52532/Keywords:
breast cancer, artificial intelligence (AI), mammography, MRI, radiomics, predictionAbstract
Relevance: Timely diagnosis of breast cancer remains one of the key challenges in healthcare, as this disease continues to be a leading cause of mortality among women worldwide. In recent years, artificial intelligence (AI) has become an integral part of medical imaging, demonstrating broad applicability and potential. Current diagnostic modalities, such as mammography and magnetic resonance imaging (MRI), serve as essential tools for detecting breast pathologies; however, they have certain limitations regarding sensitivity and specificity. This literature review presents an overview of contemporary approaches to the application of AI in the diagnosis of breast cancer.
The study aimed to analyze the methods of applying artificial intelligence in diagnosing breast cancer, including its capabilities in prediction, interpretation of results, and improving the accuracy of imaging techniques.
Methods: A comprehensive search was conducted using PubMed, Medline, Cochrane Library, and Google Scholar databases. The review includes scientific articles focused on the application of AI in the diagnosis of breast diseases.
Results: The review demonstrated that AI systems, such as convolutional neural networks, can detect microcalcifications on mammograms with high accuracy (up to 94.5%) and reduce false-positive results by 11%. In MRI image analysis, using hybrid models, such as CNN-RNN architectures, improves the diagnostic accuracy of malignant tumors by 15% and reduces error rates by 20%. Radiomics shows high accuracy (87%) in predicting therapeutic response while integrating multi-omics data provides
sensitivity up to 92%.
Conclusion: Using AI in breast cancer diagnostics enhances the accuracy of imaging techniques, facilitates data interpretation, and contributes to the personalization of treatment strategies. However, challenges remain, including the availability of high-quality data for model training and ethical considerations in decision-making processes.
