Mammographic Breast Density Model Using Semi-Supervised Learning Reduces Inter-/Intra-Reader Variability

Breast density is an important risk factor for breast cancer development; however, imager inconsistency in reporting can lead to patient and clinician confusion. A deep learning (DL) model mammographic grading was examined a retrospective multi-reader multi-case study consisting of 928 image pairs assessed impact on inter- intra-reader variability reading time. Seven readers assigned categories the images, then re-read test set aided by after 4-week washout. To measure agreement, 100 were blindly double read both sessions. Linear Cohen Kappa (κ) Student’s t-test used assess reader performance. The achieved κ 0.87 (95% CI: 0.84, 0.89) four-class assessment 0.91 0.88, 0.93) binary non-dense/dense assessment. Superiority tests showed significant reduction inter-reader (κ improved from 0.70 p ≤ 0.001) 0.83 0.95, 0.01) density, 0.77 0.96, 0.89 0.97, when DL. average mean time per pair also decreased 30%, 0.86 s 0.01, 1.71), with six seven having reductions.