An agreement has been signed for a clinical validation study involving 75,000 women: the goal is to strengthen the national prevention programme through “intelligent triage”, which optimises radiologists’ work and accelerates diagnoses.

TURIN, 12 February 2026 – With the introduction of new guidelines on AI applied to prevention at the end of 2025, Italy strengthens its role on the European stage in the fight against breast cancer, defining an advanced regulatory framework for technological integration in care pathways. Just over a month later, the Italian Mammography Screening Group (GISMa) and Italian Deep Tech company Health Triage announce the launch of an important project aimed at adding scientific evidence to the use of AI in population screening programmes through a prospective multicentre clinical study. International literature suggests that artificial intelligence can significantly contribute to improving early detection of breast cancers by supporting radiologists in highlighting suspicious areas on mammograms.

GISMa and Health Triage intend to take a further step forward and validate a reporting approach to be adopted in national screening programmes in which mammograms are triaged using Artificial Intelligence (AI).

With a cohort of 75,000 women, BREAST.AI is one of the largest Italian/European studies ever conducted using a prospective randomised design in breast imaging diagnostics. Unlike other international experiences, radiologists will work in a “blinded” mode (without knowing the score assigned by the AI): a methodological choice that strengthens the robustness and impartiality of the results, ensuring an unbiased assessment. The study therefore aims to confirm the highest diagnostic safety while optimising time and human resources.

The BREAST.AI study uses the BreastNegative software to analyse mammograms and determine their level of suspicion. Under the protocol, exams classified as “negative” by the algorithm are directed to a single human reading, while the remaining cases are read by two radiologists, as is standard in screening. This approach could reduce radiologists’ workload, allowing them to focus on more complex cases, or support an expansion of screening programme coverage. In the project, GISMa will act as an independent scientific guarantor, verifying methodological robustness, statistical analysis, and compliance with international Good Clinical Practice (GCP) standards.

“This study represents a crucial opportunity to rigorously assess the integration of artificial intelligence into screening programmes,” explains Francesca Caumo, Vice President of GISMa. “Our goal is to validate models that fully respect the principles of safety, appropriateness and quality that have always guided GISMa, ensuring that technological innovation is always in the service of women’s health and the efficiency of the National Health Service.”

“This is a decisive step to make artificial intelligence a truly useful tool and, although it is already validated in clinical practice, our goal is to gather further scientific evidence,” says Davide Dettori, CEO of Health Triage. “Our vision is clear: the goal is not to replace the physician, but to support them with advanced technologies that enhance their work. If results meet our expectations, we will finally be able to expand screening programmes, potentially extending them also to younger women thanks to the efficient use of the same human and technological resources currently available. In this perspective, the real challenge will be strengthening invitation programmes, to bring prevention to an ever-greater number of women, with particular attention to all Italian regions.”

With the entry into force of the new Guidelines of the Italian National Institute of Health (ISS), which for the first time recommend the use of AI as a triage tool to optimise screening programmes, the collaboration between GISMa and Health Triage is further reinforced. A strategic partnership that will see both parties engaged in validating the BREAST.AI study and designing new prevention models that are personalised, efficient and grounded in strong scientific evidence.