AI-assisted multi-energy mammography for improved detection in dense breasts

Challenge. Conventional 2D mammography can miss or under-characterize cancers in dense breasts due to anatomical masking. There is a need for screening-compatible solutions that improve sensitivity and/or reduce false positives while maintaining acceptable dose and workflow.

Concept. We are exploring non-contrast multi-energy mammography combined with AI fusion/decision support. The aim is to extract complementary information from multiple acquisition conditions and integrate it into outputs that are easy to use in routine practice (e.g., decision support overlays and/or derived views), while keeping the radiologist’s primary reading anchored to standard mammography images.

Objectives:

• Define clinically practical acquisition and processing requirements compatible with screening/diagnostic settings

• Develop and validate AI-assisted outputs that improve performance in dense breasts

• Establish a multi-site evaluation plan, including dataset strategy, reader studies and/or prospective feasibility work

• Address regulatory, safety, and integration requirements for clinical deployment

Partners sought:

• Clinical sites, screening programmes

• Academic imaging, medical physics groups

• Industry partners - mammography vendors / PACS / AI deployment

What we can contribute:

• AI development capability and a structured plan for clinical/technical evaluation and deployment

• A live breast-radiology collaboration platform (DICOM viewing and case workflow) and experience working with clinical partners

Collaboration interest. Open to roles as technical lead / WP lead for AI and software integration, and to co-designing a Horizon Europe Cluster 1 Health consortium around improved breast screening and diagnostic pathways.