Breast cancer is the most common malignancy in women worldwide, and mammography is the primary method of screening for the disease. Mammography has been shown to reduce mortality from the disease, but it has limitations, particularly in women with dense breasts, where the sensitivity of the method is significantly reduced. In these cases, complementary examinations such as ultrasound or magnetic resonance imaging are often necessary. The presence of false positive and negative results is a significant issue, which increases costs, benign biopsies, and psychological effects on patients.
The development of Artificial Intelligence (AI) and, in particular, deep learning algorithms (DL) offer new possibilities in the detection of cancerous lesions from mammography images. These algorithms can be trained to recognize lesions suspected of being neoplastic and provide more accurate and faster diagnosis. Although AI systems are in a pilot phase, commercially available software such as Mammoscreen is helping radiologists by reducing workload and diagnosis time by integrating them into PACS (Picture Archiving and Communication System) systems. However, it is observed that the use and combined application of these technologies in clinical routine, especially in our country, is low. This is partly due to the inherent limitations of the systems that make their acceptance unattractive in the sensitive field of medical decisions, but also to the low rate of acceptance and integration of new technologies in an environment where the traditional dominant view is the one of the experts.
The proposed research project aims to upgrade the diagnostic protocol of mammography by integrating AI. The project includes the development of an automated deep learning process using Agentic AI techniques to improve diagnostic accuracy and reduce biopsies. Various automation techniques will be studied to enable continuous training and improvement of the deep learning process. The goal is to expand the model with additional data from the University Hospital Imaging Laboratory and other data sources. The process will take into account pathological data in addition to imaging to provide a personalized application of AI to each patient. The implementation of the resulting deep learning model will be evaluated through clinical trials, aiming to demonstrate the added value of AI in clinical practice to optimize healthcare provision and the existing clinical diagnostic process.
It is also important to mention that the research results of the project will be made available and accessible through open access digital repositories, while ensuring interoperability using scientific standards. Finally, information campaigns and public actions will be carried out to disseminate the results and raise awareness about the importance of preventing breast cancer.
The overall project aims to demonstrate that the integration of these technologies can increase accuracy, reduce costs, and improve the quality of life of patients, making the new technologies more accessible to the National Health System (NHS).