Session 3: Determinants of Mammography Positioning in the MAMMO.IQ Multi-center International Study: Exploring the Impact of Compression, Breast Volume, and Age on Clinical Image Quality

Purpose: Mammography image quality is critical for accurate early breast cancer diagnosis. Mammography positioning is a primary determinant of image quality. Research highlighting individual factors such as impact of compression on positioning, is limited. A broader study of the relationship between multiple positioning errors and imaging parameters is warranted. The study aims to assess the impact of compression, breast volume, and age on screening mammography positioning.

Materials and Methods: A total of 249,817 screening mammograms from the MAMMO.IQ study were processed using A.I. algorithms to assess positioning errors on all mammograms acquired from seven breast screening programs (BSPs) from four countries across North America and Europe in the period December 1, 2019 to February 28, 2021; the study was approved by research ethics at the participating centers.

Key parameters recorded included age, breast volume, and compression pressure. Each parameter was divided into categories using deciles. Study and image-level positioning criteria included PGMI, exaggeration, portion cut off, posterior tissues missing, nipple not in profile, pectoralis shape/position, inframammary fold (IMF) missing/obscured, too high on image receptor (IR), sagging and posterior nipple line (PNL) difference.

Using the lowest category as the reference, logistic regression was used to determine odds ratios for acquisition parameters when modeled against positioning errors. The Mantel-Haenszel test assessed presence of a trend in the odds ratios.

Results: Appropriate compression can reduce many errors including CC cut off, MLO obscured, NNIP, cut off, and too high on IR. Compression is a risk factor in CC exaggeration, CC+MLO posterior tissue missing. Larger breasts are associated with reduced risk of several errors including CC exaggeration and MLO obscured. Conversely, larger breasts are associated with an increased risk of other errors including CC cut off, MLO IMF missing, and MLO sagging. Age increases risk of CC NNIP, CC cut off, and multiple MLO errors, but reduces the risk of MLO posterior tissue missing. All relationships were statistically significant (p < 0.001). No association was found between age and each of MLO sagging, BV and CC NNIP, or compression and MLO pectoralis shape, MLO IMF Missing, and CC Exaggeration.

Conclusion: This study highlights the significance of employing AI to understand the relationships between compression, breast volume, and age. By doing so, mammography positioning can be customized and optimized across diverse patient profiles, boosting diagnostic accuracy, minimizing recalls, and improving early breast cancer detection outcomes.

Clinical Relevance Statement: AI tools can enhance mammography image quality, critical in optimizing early detection of breast cancer.