Derived mammographic masking measures based on simulated lesions predict the risk of interval cancer after controlling for known risk factors: a case-case analysis

Abstract

Purpose: Women with radiographically dense or texturally complex breasts are at increased risk for interval cancer, defined as cancers diagnosed after a normal screening examination. The purpose of this study was to create masking measures and apply them to identify interval risk in a population of women who experienced either screen-detected or interval cancers after controlling for breast density.

Methods: We examined full-field digital screening mammograms acquired from 2006 to 2015. Examinations associated with 182 interval cancers were matched to 173 screen-detected cancers on age, race, exam date and time since last imaging examination. Local Image Quality Factor (IQF) values were calculated and used to create IQF maps that represented mammographic masking. We used various statistics to define global masking measures of these maps. Association of these masking measures with interval cancer vs screen-detected cancer was estimated using conditional logistic regression in a univariate and adjusted model for Breast Imaging-Reporting and Data System (BI-RADS) density. Receiver operator curves were calculated in each case to compare specificity vs sensitivity, and area under those curves were generated. Proportion of screen-detected cancer was estimated for stratifications of IQF features.

Results: Several masking features showed significant association with interval compared to screen-detected cancers after adjusting for BI-RADS density (up to P = 2.52E-6), and the 10th percentile of the IQF value (P = 1.72E-3) showed the strongest improvement in the area under the receiver operator curve, increasing from 0.65 using only BI-RADS density to 0.69. The highest masking group had a 32% proportion of screen-detected cancers while the low masking group had a 69% proportion.

Conclusions: We conclude that computer vision methods using model observers may improve quantifying the probability of breast cancer detection beyond using breast density alone.

Keywords: breast; cancer; detectability; interval; mammography; masking.

Figure 1

Sample images of an original raw data mammogram patch (left) and the same patch with a virtual lesion inserted (right) indicated by the arrow with 1 cm FWHM, 2 um Au peak thickness. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 2

Raw data mammograms (CC View) and the respective generated IQF masking maps for sample images with BI‐RADS density 1 (top‐left), 2 (top‐right), 3 (bottom‐left), and 4 (bottom‐right). Scale of IQF values are shown at right and are consistent across images. IQF values closer to zero are represented as darker pixels, indicate higher levels of masking, and are seen in the higher density images. Raw data mammograms have been contrast‐enhanced to better see dense regions, leading to an artifact being seen at the periphery. Pseudo‐presentation mammograms were developed using previously established methods within the lab [Color figure can be viewed at wileyonlinelibrary.com]

Figure 3

ROC curves for several of the masking measures in predicting interval vs screen‐detected cancer. All masking measures had similar AUCs and ROC curves in the univariate analysis. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 4

ROC curves of predicting interval vs screen‐detected cancer for the best performing masking measure (IQF 10th percentile). BI‐RADS only ROC curve has circles at each categorical cut point and lines connecting them for clarity. After controlling for BI‐RADS density, this masking measure improves the AUC from 0.65 to 0.69. P‐value for inclusion of masking measure in combined model = 1.7 E‐3. [Color figure can be viewed at wileyonlinelibrary.com]