Purpose To investigate the combination of mammography radiomics and quantitative three-compartment breast (3CB) image analysis of dual-energy mammography to limit unnecessary benign breast biopsies. Materials and Methods For this prospective study, dual-energy craniocaudal and mediolateral oblique mammograms were obtained immediately before biopsy in 109 women (mean age, 51 years; range, 31-85 years) with Breast Imaging Reporting and Data System category 4 or 5 breast masses (35 invasive cancers, 74 benign) from 2013 through 2017. The three quantitative compartments of water, lipid, and protein thickness at each pixel were calculated from the attenuation at high and low energy by using a within-image phantom. Masses were automatically segmented and features were extracted from the low-energy mammograms and the quantitative compartment images. Tenfold cross-validations using a linear discriminant classifier with predefined feature signatures helped differentiate between malignant and benign masses by means of (a) water-lipid-protein composition images alone, (b) mammography radiomics alone, and (c) a combined image analysis of both. Positive predictive value of biopsy performed (PPV₃) at maximum sensitivity was the primary performance metric, and results were compared with those for conventional diagnostic digital mammography. Results The PPV₃ for conventional diagnostic digital mammography in our data set was 32.1% (35 of 109; 95% confidence interval [CI]: 23.9%, 41.3%), with a sensitivity of 100%. In comparison, combined mammography radiomics plus quantitative 3CB image analysis had PPV₃ of 49% (34 of 70; 95% CI: 36.5%, 58.9%; P < .001), with a sensitivity of 97% (34 of 35; 95% CI: 90.3%, 100%; P < .001) and 35.8% (39 of 109) fewer total biopsies (P < .001). Conclusion Quantitative three-compartment breast image analysis of breast masses combined with mammography radiomics has the potential to reduce unnecessary breast biopsies. © RSNA, 2018 Online supplemental material is available for this article.

Neoadjuvant therapy in breast cancer can downstage axillary lymph nodes and reduce extent of axillary surgery. As such, accurate determination of nodal status after neoadjuvant therapy and before surgery impacts surgical management. There are scarce data on the diagnostic accuracy of breast magnetic resonance imaging (MRI) for nodal evaluation after neoadjuvant therapy in patients with invasive lobular carcinoma (ILC), a diffusely growing tumor type. We retrospectively analyzed patients with stage 1-3 ILC who underwent pre-operative breast MRI after either neoadjuvant chemotherapy or endocrine therapy at our institution between 2006 and 2019. Two breast radiologists reviewed MRIs and evaluated axillary nodes for suspicious features. All patients underwent either sentinel node biopsy or axillary dissection. We evaluated sensitivity, specificity, negative and positive predictive values, and overall accuracy of the post-treatment breast MRI in predicting pathologic nodal status. Of 79 patients, 58.2% received neoadjuvant chemotherapy and 41.8% neoadjuvant endocrine therapy. The sensitivity and negative predictive value of MRI were significantly higher in the neoadjuvant endocrine therapy cohort than in the neoadjuvant chemotherapy cohort (66.7 vs. 37.9%, p = 0.012 and 70.6 vs. 40%, p = 0.007, respectively), while overall accuracy was similar. Upstaging from clinically node negative to pathologically node positive occurred in 28.0 and 41.7%, respectively. In clinically node positive patients, those with an abnormal post-treatment MRI had a significantly higher proportion of patients with ≥4 positive nodes on pathology compared to those with a normal MRI (61.1 versus 16.7%, p = 0.034). Overall, accuracy of breast MRI for predicting nodal status after neoadjuvant therapy in ILC was low in both chemotherapy and endocrine therapy cohorts. However, post-treatment breast MRI may help identify patients with a high burden of nodal disease (≥4 positive nodes), which could impact pre-operative systemic therapy decisions. Further studies are needed to assess other imaging modalities to evaluate for nodal disease following neoadjuvant therapy and to improve clinical staging in patients with ILC.

Purpose To evaluate the effect of background parenchymal enhancement (BPE) on breast magnetic resonance (MR) imaging interpretive performance in a large multi-institutional cohort with independent analysis of screening and diagnostic MR studies. Materials and Methods Analysis of 3770 breast MR studies was conducted. Examinations were performed in 2958 women at six participating facilities in the San Francisco Bay Area from January 2010 to October 2012. Findings were recorded prospectively in the San Francisco Mammography Registry. Performance measures were compared between studies with low BPE (mild or minimal) and those with high BPE (moderate or marked) by using binomial tests of proportions. Results Of 1726 MR imaging studies in the screening group, 1301 were classified as having low BPE and 425 were classified as having high BPE (75% vs 25%, respectively; P < .001). Of 2044 MR imaging studies in the diagnostic group, 1443 were classified as having low BPE and 601 were classified as having high BPE (71% vs 29%, respectively; P < .001). For low versus high BPE groups at screening, abnormal interpretation rate was 157 of 1301 versus 111 of 424 (12% vs 26%, P < .001); biopsy recommendation rate was 85 of 1301 versus 54 of 424 (7% vs 13%, P < .001); and specificity was 89% (95% confidence interval [CI]: 87, 91) versus 75% (95% CI: 71, 80) (P = .01). For the low versus high BPE groups at diagnostic MR imaging, biopsy recommendation rate was 325 of 1443 versus 195 of 601 (23% vs 32%, P < .001); and specificity was 86% (95% CI: 84, 88) versus 75% (95% CI: 74, 82) (P < .001). There were no significant differences between studies with low versus high BPE in sensitivity for screening (76% [95% CI: 55, 91] vs 83% [95% CI: 52, 98]; P = .94) or diagnostic (93% [95% CI: 87, 97] vs 96% [95% CI: 87, 99]; P = .69) MR imaging, nor were there significant differences in cancer detection rate per 1000 patients between the low BPE versus high BPE groups for screening (15 per 1000 vs 24 per 1000, P = .30) or diagnostic (78 per 1000 vs 85 per 1000, P = .64) MR imaging. Conclusion Relative to MR studies with minimal or mild BPE, those with moderate or marked BPE were associated with higher abnormal interpretation and biopsy rates and lower specificity, with no difference in cancer detection rate. ^© RSNA, 2017 Online supplemental material is available for this article.

Background

To determine if mammographic features from deep learning networks can be applied in breast cancer to identify groups at interval invasive cancer risk due to masking beyond using traditional breast density measures.

Methods

Full-field digital screening mammograms acquired in our clinics between 2006 and 2015 were reviewed. Transfer learning of a deep learning network with weights initialized from ImageNet was performed to classify mammograms that were followed by an invasive interval or screen-detected cancer within 12 months of the mammogram. Hyperparameter optimization was performed and the network was visualized through saliency maps. Prediction loss and accuracy were calculated using this deep learning network. Receiver operating characteristic (ROC) curves and area under the curve (AUC) values were generated with the outcome of interval cancer using the deep learning network and compared to predictions from conditional logistic regression with errors quantified through contingency tables.

Results

Pre-cancer mammograms of 182 interval and 173 screen-detected cancers were split into training/test cases at an 80/20 ratio. Using Breast Imaging-Reporting and Data System (BI-RADS) density alone, the ability to correctly classify interval cancers was moderate (AUC = 0.65). The optimized deep learning model achieved an AUC of 0.82. Contingency table analysis showed the network was correctly classifying 75.2% of the mammograms and that incorrect classifications were slightly more common for the interval cancer mammograms. Saliency maps of each cancer case found that local information could highly drive classification of cases more than global image information.

Conclusions

Pre-cancerous mammograms contain imaging information beyond breast density that can be identified with deep learning networks to predict the probability of breast cancer detection.

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.

Ductal carcinoma in situ (DCIS) is a biologically heterogenous entity with uncertain risk for invasive ductal carcinoma (IDC) development. Standard treatment is surgical resection often followed by radiation. New approaches are needed to reduce overtreatment. This was an observational study that enrolled patients with DCIS who chose not to pursue surgical resection from 2002 to 2019 at a single academic medical center. All patients underwent breast MRI exams at 3- to 6-month intervals. Patients with hormone receptor-positive disease received endocrine therapy. Surgical resection was strongly recommended if clinical or radiographic evidence of disease progression developed. A recursive partitioning (R-PART) algorithm incorporating breast MRI features and endocrine responsiveness was used retrospectively to stratify risk of IDC. A total of 71 patients were enrolled, 2 with bilateral DCIS (73 lesions). A total of 34 (46.6%) were premenopausal, 68 (93.2%) were hormone-receptor positive, and 60 (82.1%) were intermediate- or high-grade lesions. Mean follow-up time was 8.5 years. Over half (52.1%) remained on active surveillance without evidence of IDC with mean duration of 7.4 years. Twenty patients developed IDC, of which 6 were HER2 positive. DCIS and subsequent IDC had highly concordant tumor biology. Risk of IDC was characterized by MRI features after 6 months of endocrine therapy exposure; low-, intermediate-, and high-risk groups were identified with respective IDC rates of 8.7%, 20.0%, and 68.2%. Thus, active surveillance consisting of neoadjuvant endocrine therapy and serial breast MRI may be an effective tool to risk-stratify patients with DCIS and optimally select medical or surgical management.

Significance

A retrospective analysis of 71 patients with DCIS who did not undergo upfront surgery demonstrated that breast MRI features after short-term exposure to endocrine therapy identify those at high (68.2%), intermediate (20.0%), and low risk (8.7%) of IDC. With 7.4 years mean follow-up, 52.1% of patients remain on active surveillance. A period of active surveillance offers the opportunity to risk-stratify DCIS lesions and guide decisions for operative management.

New approaches are needed to determine which ductal carcinoma in situ (DCIS) is at high risk for progression to invasive ductal carcinoma (IDC). We retrospectively studied DCIS patients who declined surgery (2002-2019), and received endocrine therapy (ET) and breast MRI. Baseline MRI and changes at 3 months and 6 months were analyzed by recursive partitioning to stratify IDC risk. Sixty-two patients (63 DCIS; 1 bilateral) with a mean follow-up of 8.5 years were included. Fifty-one percent remained on active surveillance (AS) without evidence of IDC, with a mean duration of 7.6 years. A decision tree based on MRI features of lesion distinctness and background parenchymal enhancement (BPE) at baseline and change after 3 months of ET stratified patients into low, intermediate, and high risk for progression to IDC. MRI imaging features in patients treated with ET and undergoing AS, may help determine which DCIS lesions are at low versus high risk for IDC.