Assessment of breast positioning criteria in mammographic screening: Agreement between artificial intelligence software and radiographers

Abstract

Objectives To determine the agreement between artificial intelligence software (AI) and radiographers in assessing breast positioning criteria for mammograms from standard digital mammography and digital breast tomosynthesis. Methods Assessment of breast positioning was performed by AI and by four radiographers in pairs of two on 156 examinations of women screened in Bergen, April to September 2019, as part of BreastScreen Norway. Ten criteria were used; three for craniocaudal and seven for mediolateral-oblique view. The criteria evaluated the appearance of the nipple, breast rotation, pectoral muscle, inframammary fold and pectoral nipple line. Intraclass correlation and Cohen’s kappa coefficient (κ) were used to investigate the correlation and agreement between the radiographer’s assessments and AI. Results The intraclass correlation for the pectoral nipple line between the radiographers and AI was >0.92. A substantial to almost perfect agreement (κ > 0.69) was observed between the radiographers and AI on the nipple in profile criterion. We observed a slight to moderate agreement for the other criteria (κ = 0.06–0.52) and generally a higher agreement between the two pairs of radiographers (mean κ = 0.70) than between the radiographers and AI (mean κ = 0.41). Conclusions AI has great potential in evaluating breast position criteria in mammography by reducing subjectivity. However, varying agreement between radiographers and AI was observed. Standardized and evidence-based criteria for definitions, understandings and assessment methods are needed to reach optimal image quality in mammography.

Quality Control System for Mammographic Breast Positioning Using Deep Learning

“…Further, an image evaluation method using a DCNN for skull X-ray images [22] has been reported. The most recent attempts to evaluate mammographic breast criteria have not clari ed the quantitative values derived from traditional image processing techniques [23] . We believe that our study is the rst to address mammography breast positioning using DCNN classi cation.…”

“…Deep learning has also been applied to positioning for X-ray examinations [21,22] . Although there are several approaches to assessing mammographic positioning using computer schemes [23] , to the best of our knowledge, there are no reports on the use of DCNN for the veri cation of optimal positioning. In this study, we propose a DCNN classi cation for the quality control and validation of positioning in mammography, in which each part of a mammogram can be detected automatically.…”

Quantitative Breast Density in Contrast-Enhanced Mammography

“…For CEM examinations, VBD computing was limited to the LE images. To help explain potential sources of density variability, other parameters studied for their potential association with VBD are breast thickness and compression force (both obtained from the image DICOM header), area of the compression paddle in contact with the breast [ 37 ], distance between the nipple and the posterior-edge (both determined by image analysis) [ 38 ], compression pressure calculated as the ratio between compression force and contact area, and mean glandular dose (MGD) obtained by applying the dosimetry model proposed by Dance and colleagues [ 39 , 40 , 41 ].…”

Standardised or individualised X-ray tube angle for mediolateral oblique projection in digital mammography?

“…Volpara is developed to measure image quality as one measure, where the pectoral muscle appearance represents one out of several measures (16,22). However, a previous study made quantitative measurements of pectoral muscle angle and length to posterior nipple line and compared human versus Volpara metrics (23). That study showed adequate accuracy of the Volpara measures, especially to the level needed to observe trends within large sample sizes.…”