AI & Quality management

Team: Widmann G, Santner T, Galijasevic M, Gruber L, Amort B

A PhD project (Tina Santner, MSc.) on quality, safety and standards in mammography is dedicated to assessing the diagnostic image quality, in particular the correct positioning of the breast in the various images. For this purpose, an international multi-reader study was initiated in which five different mammography experts carry out a PGMI image quality assessment (perfect, good, moderate, insufficient) with mammography data sets from Austria, Norway and Switzerland and compare their results with each other and with the assessment of available AI software. The aim is to develop the basis for the use of AI software in screening programs and for routine mammography, with the function of immediate feedback on image quality as well as continuous quality monitoring.
A joint ESR (European Society of Radiology) Quality Safety and Standards Committee and EUSOBI (European Society of Breast Imaging) project has been initiated to establish standardised criteria for evaluating mammographies in Europe. Goal is to define a European standard for measuring, evaluating, documenting, and utilizing artificial intelligence tools in quality assessment effectively.

Abb. 1. Dashboard of an AI-based PGMI analysis of a mammography study data set.

Abb. 2. AI-based analysis: CC views show pectoralis muscle (although shorter PNL than MLO), nipple in profile, parenchyma captured with "perfect" rating. MLO views do not show visualization of the IMF (inframammary fold), nipple LMLO not in profile with rating "moderate".

Selected publication
Image quality assessment in screening mammography: How subjective is PGMI really and can AI software help? Santner T, Ruppert C, Gianolini S, Stalheim J-G, Frei S, Hondl Adametz M, Fröhlich V, Hofvind S, Widmann G. EPOS C-19422. ECR 2024. Vienna, Austria March 2024.