MammoScreen® (4)

MammoScreen 4 is an AI-powered software device that assists radiologists in interpreting mammograms by detecting and characterizing breast lesions suspicious for cancer. It processes full-field digital mammography and digital breast tomosynthesis images, marks detected findings with suspicion scores (1-10), classifies lesions as masses/asymmetries, distortions, or calcifications, and provides location information. The device is intended as a concurrent reading aid to support physician decision-making, not replace clinical judgment.

MammoScreen 4 uses the same machine learning-based AI architecture as its predicate (MammoScreen 3) and maintains identical outputs: graphical marks, suspicion scores at finding/breast/case levels, lesion type classification, and spatial localization. The main differences are support for Hologic Envision systems (new in version 4) and inclusion of a Predetermined Change Control Plan for future algorithm modifications. Both versions process 2D and 3D mammography with optional prior examinations.

Testing complied with IEC 62304:2006/A1:2016 (Medical device software - Software life-cycle processes) and IEC 62366-1:2015+AMD1:2020 (Medical devices - Application of usability engineering to medical devices). Standalone performance validation included 1,475 patients and 2,950 studies with non-inferiority testing. Clinical validation comprised three multi-reader multi-case (MRMC) studies for FFDM, DBT, and combined modalities.

MammoScreen 4 is substantially equivalent to MammoScreen 3 because both devices have identical intended use (concurrent reading aid for mammogram interpretation), same fundamental AI/machine learning technology, identical output formats (suspicion scores, lesion characterization, localization), and serve the same user and patient populations. The design changes (extended system support) do not raise different safety or effectiveness questions. Non-inferiority standalone testing demonstrated superior performance compared to the predicate, and MRMC studies showed radiologists using MammoScreen 4 perform better than unaided radiologists, establishing safety and effectiveness.