Patient Preferences of Breast Radiologists' Attire and Appearance.

RATIONALE AND OBJECTIVES: To assess patient preferences for breast radiologists' attire and appearance. MATERIALS AND METHODS: A multi-institutional anonymous, voluntary 19-question survey was administered to patients undergoing screening and diagnostic mammography examinations over a 5-week period. Using a 5-point Likert scale, respondents were asked about their preferences for gender-neutral attire (white coat), male-presenting attire (scrubs, dress shirt with tie, or dress shirt without tie), and female-presenting attire (scrubs, dress, blouse with pants, and blouse with skirt). Patient responses were compared to demographic data using bivariable analysis and multivariable regression. RESULTS: Response rate was 84.7% (957/1130). Mean respondent age was 57.2 years±11.9. Most respondents agreed/strongly agreed that the breast radiologist's appearance mattered (52.5%, 502/956) followed by being indifferent (28.1%, 269/956). Respondents with greater education levels felt less strongly (p=0.001) about the radiologist's appearance: 63.3% (70/110) less than college cared about appearance compared to 53.5% (266/497) college/vocational and 47.4% (165/348) graduate. Most respondents felt indifferent about a breast radiologist wearing a white coat (68.9%, 657/954) or about male-presenting breast radiologists wearing a tie (77.1%, 734/952) without significant demographic differences. Almost all respondents either prefer/strongly prefer (60.1%, 572/951) or were indifferent (39.6%, 377/951) to all breast radiologists wearing scrubs when performing procedures. While respondents approved of all attire choices overall, most respondents preferred scrubs for both male- and female-presenting breast radiologists (64.0%, 612/957 and 64.9%, 621/957, respectively). CONCLUSION: A variety of breast radiologists' attire can be worn while maintaining provider professionalism and without compromising patient expectations.

Impact of Interruptions During Screening Mammography on Physician Well-Being and Patient Care.

OBJECTIVE: To assess the impact of interruptions on radiologists' efficiency, accuracy, and job satisfaction in interpreting screening mammograms. METHODS: This institutional review board-approved retrospective reader study recruited nine breast radiologists from a single academic institution [name withheld] to interpret 150 screening mammograms performed between December 1, 2008, and December 31, 2015 under two different reading conditions, as follows: (1) uninterrupted batch reading and (2) interrupted reading. The 150 cases consisted of 125 normal mammograms and 25 mammograms with subtle breast cancers. Cases were divided into two groups of 75 cases each (cohort 1 and cohort 2), with a comparable distribution of cancer cases. Four rounds of 75 cases each were conducted with a 6-week washout period between rounds 2 and 3. After completing each interpretation session, readers completed a seven-question survey, assessing perceptions of mental and physical effort, level of frustration, and performance satisfaction. Clinical performance metrics (reading time, recall rate, sensitivity, specificity, accuracy, and positive predictive value 1) were calculated. RESULTS: Recall rates were significantly (P = .04) higher during interrupted reading sessions (35.4%) than they were during uninterrupted batch reading sessions (31.4%). Accuracy was significantly (P = .049) worse in the interrupted reading sessions (69.5%), compared with uninterrupted sessions (73.6%). Differences in overall image interpretation times were not statistically significant (P = .065). Compared with uninterrupted batch reading sessions, readers during interrupted sessions reported feeling busier (P < .001), encountered higher levels of cognitive demand (P = .005), experienced elevated levels of physical fatigue (P = .004), and expressed lower levels of satisfaction with their performance (P = .041). CONCLUSION: Interruptions during interpretation of screening mammography have deleterious effects on physician performance and their sense of well-being.

Breast Cancer Risk Assessment and Screening Practices Reported Via an Online Survey.

BACKGROUND: Breast cancer screening guidelines differ between organizations, and significant variations in practice patterns exist. Previous evidence suggests that provider-level factors are the greatest contributors to risk assessment and screening practice variability. This study aimed to characterize provider factors associated with breast cancer risk assessment and screening practice patterns, and to assess perceived barriers to providing risk assessment. METHODS: An online survey was distributed to providers at a single academic institution and to providers publicly via social media (January to August 2022). Respondents in the United States who care for adult women at risk for the development of breast cancer were included. RESULTS: Most of the respondents in the 143 completed surveys were white/Caucasian (79%) females (90%) age 50 years or younger (79%), and whereas 97% discuss breast cancer screening with their patients, only 90% order screening mammograms. Risk factor assessment was common (93%), typically performed at the first visit (51%). Additional training in genetics or risk assessment was uncommon (17%), although the majority were interested but did not have the time or resources (55%). Although most (64%) did not perceive barriers to providing risk assessment or appropriate screening, the most common barriers were time (77%) and education (55%). Barriers were more common among family practice or obstetrics and gynecology (OB/GYN) providers and those who worked in an academic setting (all p < 0.05). CONCLUSIONS: Breast cancer risk assessment and screening practices are highly variable. Although time is the major barrier to providing risk assessment, providers also need education. Primary care organizations could partner with breast cancer-focused societies for additional resources.

Contrast-Enhanced In-Phase Dixon Sequence: Impact on Biopsy Clip Detection on Breast MRI.

BACKGROUND. Identification of breast biopsy clips using conventional MRI sequences may be challenging. A contrast-enhanced in-phase Dixon sequence may have greater conspicuity for areas of susceptibility compared with standard clinical sequences. OBJECTIVE. The purpose of this article is to compare detection of breast biopsy clips on MRI between the contrast-enhanced in-phase Dixon sequence and three routine clinical sequences. METHODS. This retrospective study included 164 patients (mean age, 50.3 years) with a total of 281 breast biopsy clips who underwent contrast-enhanced breast MRI between January 2, 2019, and April 16, 2020. Three radiologists, blinded to the clip location and sequence used, independently annotated biopsy clip locations on three clinical sequences (T1-weighted non-fat-suppressed [NFS], STIR, and first phase from dynamic contrast-enhanced T1-weighted fat-suppressed [FS]) and on a contrast-enhanced in-phase Dixon sequence and then recorded confidence scores (1-4 scale). A study coordinator used all available imaging and reports to localize clips on MRI, which served as the reference standard. A physicist measured clip CNR. Sequences were compared using the McNemar test and two-tailed Wilcoxon signed rank tests. RESULTS. Among the three readers, pooled sensitivity and PPV were 78.2% and 96.2% for T1-weighted NFS, 26.6% and 92.7% for STIR, 61.7% and 95.9% for contrast-enhanced T1-weighted FS, and 85.1% and 95.1% for contrast-enhanced in-phase Dixon sequence. Pooled sensitivity was higher for contrast-enhanced in-phase Dixon sequence than for the other sequences (all p < .05); pooled PPV was not significantly different between contrast-enhanced in-phase Dixon and the other sequences (all p > .05). Mean confidence scores (pooled across readers for true-positive assessments) and mean CNR were 3.0 ± 0.9 (SD) and 1.21 ± 0.61 for T1-weighted NFS, 1.7 ± 0.9 and 0.57 ± 0.69 for STIR, 2.5 ± 1.0 and 0.54 ± 0.61 for contrast-enhanced T1-weighted FS, and 3.5 ± 0.8 and 4.05 ± 2.6 for the contrast-enhanced in-phase Dixon sequence. Pooled mean confidence scores and CNR were higher for contrast-enhanced in-phase Dixon than for the other sequences (all p < .001). CONCLUSION. Compared with clinical sequences, the contrast-enhanced in-phase Dixon sequence had higher sensitivity for detecting breast biopsy clips on MRI and higher reader confidence and CNR, without change in PPV. CLINICAL IMPACT. The contrast-enhanced in-phase Dixon sequence may help address a current challenge in clinical breast MRI interpretation.

Ethanol Sclerotherapy for Postoperative Seroma of the Breast and Axilla.

OBJECTIVE: Evaluate the effectiveness of alcohol sclerotherapy in postoperative breast and axillary seromas. METHODS: This was an IRB-approved retrospective review of consecutive patients from 2017 to 2021. The procedure involves aspiration of seroma fluid, injection of ethanol for 15 to 30 minutes, and then aspiration of the injected ethanol. Following review of the medical record, patient and procedure data were recorded. Success was defined as no recurrence of seroma. Statistical analysis was performed using a chi-square or t-test, as appropriate. RESULTS: Twenty seromas were treated in 19 patients (mean age, 63 years; range, 49-79) following mastectomy (9/20, 45%), lumpectomy (25%, 5/20), axillary nodal dissection (5/20, 25%), or chest wall surgery (1/20, 5%). The mean seroma volume was 146 ±â€…165 mL. There was a mean of 1.4 prior aspiration attempts (range, 0-3). Mean injected ethanol volume was 39 mL (range, 8-60 mL) for a mean of 26 ±â€…5 minutes. Seromas recurred in 85% (17/20) of cases, and repeat aspiration was performed in 35% (6/17). The mean recurrence time was 34 ±â€…29 days. The recurrent seromas were 31% smaller (mean, 80 ±â€…48 mL). Only a smaller pre-ablation seroma was associated with successful ablation (P < 0.01), as the mean pre-ablation volume was 36 ± 24 mL for successful ablations and 119 ±â€…56 mL for unsuccessful ablations. There were no post-procedural complications. CONCLUSION: Single-treatment ethanol sclerotherapy is typically unsuccessful in ablating breast and axillary seromas. Only a smaller pre-ablation seroma size was associated with successful ablation.

Benefits and Risks of Mammography Screening in Women Ages 40 to 49 Years.

Breast cancer screening in the United States is complicated by conflicting recommendations from professional and governmental organizations. The benefits and risks of breast cancer screening differ though by age which should influence shared decision-making discussions. Compared to older women, women ages 40 to 49 years have a lower risk of breast cancer, but the types of breast cancer that develop are often more aggressive with a poorer prognosis. Furthermore, younger women have a longer life expectancy and fewer comorbidities. The primary benefits of screening for women in their 40s are a reduction in breast cancer mortality, years of life lost to breast cancer, and morbidity of breast cancer treatment by detecting cancers at an earlier stage. Compared to older women, the risks of breast cancer screening in women ages 40 to 49 years includes more false positive recalls and biopsies as well as transient anxiety. Concerns regarding radiation induced malignancy and overdiagnosis are minimal in this age group. The shorter lead time of breast cancer in women ages 40 to 49 years also favors shorter screening intervals. This information should help inform providers in their shared decision-making discussions with patients.

Mixed-Methods Study to Predict Upstaging of DCIS to Invasive Disease on Mammography.

BACKGROUND. The incidence of ductal carcinoma in situ (DCIS) has steadily increased, as have concerns regarding overtreatment. Active surveillance is a novel treatment strategy that avoids surgical excision, but identifying patients with occult invasive disease who should be excluded from active surveillance is challenging. Radiologists are not typically expected to predict the upstaging of DCIS to invasive disease, though they might be trained to perform this task. OBJECTIVE. The purpose of this study was to determine whether a mixed-methods two-stage observer study can improve radiologists' ability to predict upstaging of DCIS to invasive disease on mammography. METHODS. All cases of DCIS calcifications that underwent stereotactic biopsy between 2010 and 2015 were identified. Two cohorts were randomly generated, each containing 150 cases (120 pure DCIS cases and 30 DCIS cases upstaged to invasive disease at surgery). Nine breast radiologists reviewed the mammograms in the first cohort in a blinded fashion and scored the probability of upstaging to invasive disease. The radiologists then reviewed the cases and results collectively in a focus group to develop consensus criteria that could improve their ability to predict upstaging. The radiologists reviewed the mammograms from the second cohort in a blinded fashion and again scored the probability of upstaging. Statistical analysis compared the performances between rounds 1 and 2. RESULTS. The mean AUC for reader performance in predicting upstaging in round 1 was 0.623 (range, 0.514-0.684). In the focus group, radiologists agreed that upstaging was better predicted when an associated mass, asymmetry, or architectural distortion was present; when densely packed calcifications extended over a larger area; and when the most suspicious features were focused on rather than the most common features. Additionally, radiologists agreed that BI-RADS descriptors do not adequately characterize risk of invasion, and that microinvasive disease and smaller areas of DCIS will have poor prediction estimates. Reader performance significantly improved in round 2 (mean AUC, 0.765; range, 0.617-0.852; p = .045). CONCLUSION. A mixed-methods two-stage observer study identified factors that helped radiologists significantly improve their ability to predict upstaging of DCIS to invasive disease. CLINICAL IMPACT. Breast radiologists can be trained to better predict upstaging of DCIS to invasive disease, which may facilitate discussions with patients and referring providers.

Predicting Upstaging of DCIS to Invasive Disease: Radiologists's Predictive Performance.

RATIONALE AND OBJECTIVES: The purpose of this study is to quantify breast radiologists' performance at predicting occult invasive disease when ductal carcinoma in situ (DCIS) presents as calcifications on mammography and to identify imaging and histopathological features that are associated with radiologists' performance. MATERIALS AND METHODS: Mammographically detected calcifications that were initially diagnosed as DCIS on core biopsy and underwent definitive surgical excision between 2010 and 2015 were identified. Thirty cases of suspicious calcifications upstaged to invasive ductal carcinoma and 120 cases of DCIS confirmed at the time of definitive surgery were randomly selected. Nuclear grade, estrogen and progesterone receptor status, patient age, calcification long axis length, and breast density were collected. Ten breast radiologists who were blinded to all clinical and pathology data independently reviewed all cases and estimated the likelihood that the DCIS would be upstaged to invasive disease at surgical excision. Subgroup analysis was performed based on nuclear grade, long axis length, breast density and after exclusion of microinvasive disease. RESULTS: Reader performance to predict upstaging ranged from an area under the receiver operating characteristic curve (AUC) of 0.541-0.684 with a mean AUC of 0.620 (95%CI: 0.489-0.751). Performances improved for lesions smaller than 2 cm (AUC: 0.676 vs 0.500; p = 0.002). The exclusion of microinvasive cases also improved performance (AUC: 0.651 vs 0.620; p = 0.005). There was no difference in performance based on breast density (p = 0.850) or nuclear grade (p = 0.270) CONCLUSION: Radiologists were able to predict invasive disease better than chance, particularly for smaller DCIS lesions (<2 cm) and after the exclusion of microinvasive disease.

Impact of Colorized Display of Mammograms on Lesion Detection.

OBJECTIVE: To assess the effect of the colorized display of digital mammograms on observer detection of subtle breast lesions. METHODS: Three separate observer studies compared detection performance using grayscale versus color display of 1) low-contrast mass-like objects in a standardized mammography phantom; 2) simulated microcalcifications in a background of normal breast parenchyma; and 3) standard-of-care clinical digital mammograms with subtle calcifications and masses. Colorization of the images was done by displaying each image pixel in blue, green, and red hues, or gray, maintaining DICOM-calibrated luminance scale and consistent luminance range. For the simulated calcifications and clinical mammogram studies, comparison of detection rates was computed using McNemar's test for paired differences. RESULTS: For the phantom study, mass-like object detection was significantly better using a green colormap than grayscale (73.3% vs 70.8%, P = .009), with no significant improvement using blue or red colormaps (72.6% and 72.5%, respectively). For simulated microcalcifications, no significant difference was noted in detection using the green colormap, as compared with grayscale. For clinical digital screening mammograms, no significant difference was noted between gray and green colormaps for detection of microcalcifications. Green color display, however, resulted in decreased sensitivity for detection of subtle masses (63% vs 69%, P = .03). CONCLUSION: Although modest improvement was demonstrated for a detection task using colorized display of a standard mammography phantom, no significant improvement was demonstrated using a color display for a simulated clinical detection task, and actual clinical performance was worse for colorized display of mammograms in comparison to standard grayscale display.

MRI predictors of tumor-positive margins after breast-conserving surgery.

PURPOSE: The purpose of this study is to identify predictors of tumor-positive surgical margins after breast-conserving surgery on dynamic contrast-enhanced (DCE) MRI. MATERIALS AND METHODS: We conducted a retrospective study of consecutive women who underwent DCE MRI before breast-conserving surgery from 2005 to 2014. Patient demographics, indication for surgery, MRI findings, biopsy pathology results, and surgical outcomes were reviewed. The unpaired t-test and chi-square test were used to compare the positive and negative margins groups. RESULTS: 554 women (mean age, 56; range, 26-90) underwent DCE MRI before 575 breast-conserving surgeries for invasive carcinoma (n = 473) or ductal carcinoma in situ (DCIS) (n = 102). Positive margins requiring re-excision occurred in 19.7% (93/473) of surgeries for invasive carcinoma and 31.4% (32/102) of surgeries for DCIS. For invasive carcinoma surgeries, positive margins were more common when MRI demonstrated the finding of non-mass enhancement (NME) rather than the finding of enhancing mass (33.8% [22/65] versus 16.9% [61/360], p < 0.01). Tumor size on MRI was significantly larger in the positive margins group (2.5 cm versus 1.9 cm, p < 0.001). Positive margins were more common with invasive lobular rather than invasive ductal histology at core biopsy (38.3% [18/47] versus 16.0% [56/350], p < 0.001). For DCIS surgeries, there were no significant differences in positive margin rates related to MRI features. CONCLUSION: For invasive carcinoma surgeries, positive margins are associated with NME on MRI, larger tumor size on MRI, and lobular histology at core biopsy. These findings may be used to predict which patients are at risk for positive margins after breast-conserving surgery.

Can Digital Breast Tomosynthesis Replace Full-Field Digital Mammography? A Multireader, Multicase Study of Wide-Angle Tomosynthesis.

OBJECTIVE. The purpose of this study was to test the hypothesis whether two-view wide-angle digital breast tomosynthesis (DBT) can replace full-field digital mammography (FFDM) for breast cancer detection. SUBJECTS AND METHODS. In a multireader multicase study, bilateral two-view FFDM and bilateral two-view wide-angle DBT images were independently viewed for breast cancer detection in two reading sessions separated by more than 1 month. From a pool of 764 patients undergoing screening and diagnostic mammography, 330 patient-cases were selected. The endpoints were the mean ROC AUC for the reader per breast (breast level), ROC AUC per patient (subject level), noncancer recall rates, sensitivity, and specificity. RESULTS. Twenty-nine of 31 readers performed better with DBT than FFDM regardless of breast density. There was a statistically significant improvement in readers' mean diagnostic accuracy with DBT. The subject-level AUC increased from 0.765 (standard error [SE], 0.027) for FFDM to 0.835 (SE, 0.027) for DBT (p = 0.002). Breast-level AUC increased from 0.818 (SE, 0.019) for FFDM to 0.861 (SE, 0.019) for DBT (p = 0.011). The noncancer recall rate per patient was reduced by 19% with DBT (p < 0.001). Masses and architectural distortions were detected more with DBT (p < 0.001); calcifications trended lower (p = 0.136). Accuracy for detection of invasive cancers was significantly greater with DBT (p < 0.001). CONCLUSION. Reader performance in breast cancer detection is significantly higher with wide-angle two-view DBT independent of FFDM, verifying the robustness of DBT as a sole view. However, results of perception studies in the vision sciences support the inclusion of an overview image.

Breast cancer deaths averted over 3 decades.

BACKGROUND: From 1975 to 1990, female breast cancer mortality rates in the United States increased by 0.4% per year. Since 1990, breast cancer mortality rates have fallen between 1.8% and 3.4% per year, a decrease that is attributed to increased mammography screening and improved treatment. METHODS: The authors used age-adjusted female breast cancer mortality rate and population data from the Surveillance, Epidemiology, and End Results (SEER) program to estimate the number of breast cancer deaths averted by screening mammography and improved treatment since 1989. Four different assumptions regarding background mortality rates (in the absence of screening mammography and improved treatment) were used to estimate deaths averted for women aged 40 to 84 years by taking the difference between SEER-reported mortality rates and background mortality rates for each 5-year age group, multiplied by the population for each 5-year age group. SEER data were used to estimate annual and cumulative breast cancer deaths averted in 2012 and 2015 and extrapolated SEER data were used to estimate deaths averted in 2018. RESULTS: The number of single-year breast cancer deaths averted ranged from 20,860 to 33,842 in 2012, from 23,703 to 39,415 in 2015, and from 27,083 to 45,726 in 2018. Breast cancer mortality reductions ranged from 38.6% to 50.5% in 2012, from 41.5% to 54.2% in 2015, and from 45.3% to 58.3% in 2018. Cumulative breast cancer deaths averted since 1989 ranged from 237,234 to 370,402 in 2012, from 305,934 to 483,435 in 2015, and from 384,046 to 614,484 in 2018. CONCLUSIONS: Since 1989, between 384,000 and 614,500 breast cancer deaths have been averted through the use of mammography screening and improved treatment.

A Proposal to Define Three New Breast Calcification Shapes: Square, Sandwich, and Teardrop, Pill & Capsule.

OBJECTIVE: To define and illustrate three new descriptors for microcalcification morphology on mammography, and provide a preliminary assessment of their potential clinical value. METHODS: The institutional review board approved this retrospective study and waived the requirement for informed consent. One thousand consecutive biopsy-proven cases of microcalcifications were retrospectively reviewed by the index study observer, a fellowship-trained breast radiologist blinded to histologic outcome. Each case was evaluated for traditional Breast Imaging Reporting and Data System (BI-RADS) morphology descriptors, as well as for the presence of one of three proposed new calcification morphology descriptors: (1) square, (2) sandwich, and (3) teardrop, pill & capsule. Positive predictive value (PPV) for each proposed shape was calculated. Interobserver variability was calculated on a subset of 200 cases, which were evaluated by two additional breast radiologists blinded to the interpretation of the index observer. RESULTS: Of 1000 cases, 702 (70.2%) were benign, and 298 (29.8%) were malignant. There were 51 (5.1%) square and 60 (6.0%) sandwich calcification cases, which were all benign at biopsy (PPV, 0%). Of the 106 (10.6%) cases that included teardrop, pill & capsule calcifications, 90 were malignant (PPV, 84.9%). Interobserver variability for the traditional BI-RADS morphology descriptors (κ = 0.40) was comparable to observer variability of the proposed new morphologies: square (κ = 0.39), sandwich (κ = 0.47), and teardrop, pill & capsule (κ = 0.49). CONCLUSION: The proposed calcification descriptors square and sandwich-shaped had a PPV of 0%, and they may represent new typically benign morphologies. In contrast, teardrop, pill & capsule-shaped calcifications indicate a higher likelihood of malignancy than that of previously established descriptors.

Findings on Surveillance Imaging After Preoperative Partial Breast Irradiation for Early Stage Breast Cancer.

PURPOSE: To evaluate the mammographic sequelae of preoperative accelerated partial breast irradiation (APBI) delivered via either stereotactic radiosurgery or a conventionally fractionated regimen. METHODS AND MATERIALS: This multicenter, retrospective study evaluated surveillance mammograms from patients enrolled in 2 prospective, preoperative APBI clinical trials. At 1 site, 31 patients with cT1N0 invasive carcinomas or low- or intermediate-grade ductal carcinoma in situ (<2 cm) received preoperative stereotactic radiosurgery and had a total of 186 mammograms available for review. At the second site, 180 mammograms from 25 patients with cT1-2 (<3 cm) unifocal invasive carcinomas treated with conventionally fractionated, preoperative APBI were reviewed. Findings were compared with those of 26 early stage breast cancers treated with conventional postoperative whole breast radiation therapy. RESULTS: At a median follow-up of 61 months, 17 patients (55%) treated with single-dose APBI exhibited exuberant fat necrosis at the lumpectomy site. Fat necrosis was believed to be clinically palpable in 5 (16%) of these patients within the first 3 years of follow-up. Exuberant fat necrosis developed in 5 patients (20%) treated with fractionated APBI over a median 68-month follow-up period but only 2 of those patients (8%) who underwent conventional whole breast radiation therapy. CONCLUSIONS: In situ tumor targeting in the preoperative setting allows relative sparing of normal tissue but results in a larger and more vigorous area of change on surveillance imaging, potentially reflecting the interaction of surgical resection with an irradiated tissue bed. High-dose stereotactic radiosurgery in particular increases the risk of developing a uniquely robust and well-demarcated pattern of fat necrosis on mammogram that may also present clinically. With many ongoing studies evaluating the preoperative treatment approach, defining the landscape of expected imaging sequelae will provide useful anticipatory guidance for clinicians and patients.

Prediction of Occult Invasive Disease in Ductal Carcinoma in Situ Using Deep Learning Features.

PURPOSE: The aim of this study was to determine whether deep features extracted from digital mammograms using a pretrained deep convolutional neural network are prognostic of occult invasive disease for patients with ductal carcinoma in situ (DCIS) on core needle biopsy. METHODS: In this retrospective study, digital mammographic magnification views were collected for 99 subjects with DCIS at biopsy, 25 of which were subsequently upstaged to invasive cancer. A deep convolutional neural network model that was pretrained on nonmedical images (eg, animals, plants, instruments) was used as the feature extractor. Through a statistical pooling strategy, deep features were extracted at different levels of convolutional layers from the lesion areas, without sacrificing the original resolution or distorting the underlying topology. A multivariate classifier was then trained to predict which tumors contain occult invasive disease. This was compared with the performance of traditional "handcrafted" computer vision (CV) features previously developed specifically to assess mammographic calcifications. The generalization performance was assessed using Monte Carlo cross-validation and receiver operating characteristic curve analysis. RESULTS: Deep features were able to distinguish DCIS with occult invasion from pure DCIS, with an area under the receiver operating characteristic curve of 0.70 (95% confidence interval, 0.68-0.73). This performance was comparable with the handcrafted CV features (area under the curve = 0.68; 95% confidence interval, 0.66-0.71) that were designed with prior domain knowledge. CONCLUSIONS: Despite being pretrained on only nonmedical images, the deep features extracted from digital mammograms demonstrated comparable performance with handcrafted CV features for the challenging task of predicting DCIS upstaging.

Can Occult Invasive Disease in Ductal Carcinoma In Situ Be Predicted Using Computer-extracted Mammographic Features?

RATIONALE AND OBJECTIVES: This study aimed to determine whether mammographic features assessed by radiologists and using computer algorithms are prognostic of occult invasive disease for patients showing ductal carcinoma in situ (DCIS) only in core biopsy. MATERIALS AND METHODS: In this retrospective study, we analyzed data from 99 subjects with DCIS (74 pure DCIS, 25 DCIS with occult invasion). We developed a computer-vision algorithm capable of extracting 113 features from magnification views in mammograms and combining these features to predict whether a DCIS case will be upstaged to invasive cancer at the time of definitive surgery. In comparison, we also built predictive models based on physician-interpreted features, which included histologic features extracted from biopsy reports and Breast Imaging Reporting and Data System-related mammographic features assessed by two radiologists. The generalization performance was assessed using leave-one-out cross validation with the receiver operating characteristic curve analysis. RESULTS: Using the computer-extracted mammographic features, the multivariate classifier was able to distinguish DCIS with occult invasion from pure DCIS, with an area under the curve for receiver operating characteristic equal to 0.70 (95% confidence interval: 0.59-0.81). The physician-interpreted features including histologic features and Breast Imaging Reporting and Data System-related mammographic features assessed by two radiologists showed mixed results, and only one radiologist's subjective assessment was predictive, with an area under the curve for receiver operating characteristic equal to 0.68 (95% confidence interval: 0.57-0.81). CONCLUSIONS: Predicting upstaging for DCIS based upon mammograms is challenging, and there exists significant interobserver variability among radiologists. However, the proposed computer-extracted mammographic features are promising for the prediction of occult invasion in DCIS.