Faculty Bibliography

Machine-learning techniques have led to remarkable advances in data extraction and analysis of medical imaging. Applications of machine learning to breast MRI continue to expand rapidly as increasingly accurate 3D breast and lesion segmentation allows the combination of radiologist-level interpretation (eg, BI-RADS lexicon), data from advanced multiparametric imaging techniques, and patient-level data such as genetic risk markers. Advances in breast MRI feature extraction have led to rapid dataset analysis, which offers promise in large pooled multiinstitutional data analysis. The object of this review is to provide an overview of machine-learning and deep-learning techniques for breast MRI, including supervised and unsupervised methods, anatomic breast segmentation, and lesion segmentation. Finally, it explores the role of machine learning, current limitations, and future applications to texture analysis, radiomics, and radiogenomics. Level of Evidence: 3 Technical Efficacy Stage: 2 J. Magn. Reson. Imaging 2019.

Breast MRI offers high sensitivity for breast cancer detection, with preferential detection of high-grade invasive cancers when compared to mammography and ultrasound. Despite the clear benefits of breast MRI in cancer screening, its cost, patient tolerance, and low utilization remain key issues. Abbreviated breast MRI, in which only a select number of sequences and postcontrast imaging are acquired, exploits the high sensitivity of breast MRI while reducing table time and reading time to maximize availability, patient tolerance, and accessibility. Worldwide studies of varying patient populations have demonstrated that the comparable diagnostic accuracy of abbreviated breast MRI is comparable to a full diagnostic protocol, highlighting the emerging role of abbreviated MRI screening in patients with an intermediate and high lifetime risk of breast cancer. The purpose of this review is to summarize the background and current literature relating to abbreviated MRI, highlight various protocols utilized in current multicenter clinical trials, describe workflow and clinical implementation issues, and discuss the future of abbreviated protocols, including advanced MRI techniques.

Breast MRI offers high sensitivity for breast cancer detection, with preferential detection of high-grade invasive cancers when compared to mammography and ultrasound. Despite the clear benefits of breast MRI in cancer screening, its cost, patient tolerance, and low utilization remain key issues. Abbreviated breast MRI, in which only a select number of sequences and postcontrast imaging are acquired, exploits the high sensitivity of breast MRI while reducing table time and reading time to maximize availability, patient tolerance, and accessibility. Worldwide studies of varying patient populations have demonstrated that the comparable diagnostic accuracy of abbreviated breast MRI is comparable to a full diagnostic protocol, highlighting the emerging role of abbreviated MRI screening in patients with an intermediate and high lifetime risk of breast cancer. The purpose of this review is to summarize the background and current literature relating to abbreviated MRI, highlight various protocols utilized in current multicenter clinical trials, describe workflow and clinical implementation issues, and discuss the future of abbreviated protocols, including advanced MRI techniques.

We present a deep convolutional neural network for breast cancer screening exam classification, trained and evaluated on over 200,000 exams (over 1,000,000 images). Our network achieves an AUC of 0.895 in predicting the presence of cancer in the breast, when tested on the screening population. We attribute the high accuracy to a few technical advances. (i) Our network's novel two-stage architecture and training procedure, which allows us to use a high-capacity patch-level network to learn from pixel-level labels alongside a network learning from macroscopic breast-level labels. (ii) A custom ResNet-based network used as a building block of our model, whose balance of depth and width is optimized for high-resolution medical images. (iii) Pretraining the network on screening BI-RADS classification, a related task with more noisy labels. (iv) Combining multiple input views in an optimal way among a number of possible choices. To validate our model, we conducted a reader study with 14 readers, each reading 720 screening mammogram exams, and show that our model is as accurate as experienced radiologists when presented with the same data. We also show that a hybrid model, averaging the probability of malignancy predicted by a radiologist with a prediction of our neural network, is more accurate than either of the two separately. To further understand our results, we conduct a thorough analysis of our network's performance on different subpopulations of the screening population, the model's design, training procedure, errors, and properties of its internal representations. Our best models are publicly available at https://github.com/nyukat/breastcancerclassifier.

Vascular lesions (VLs) of the breast present a diagnostic challenge on breast core biopsy (BCBx). We report on 27 VLs presenting on BCBx. The mean patient age was 60 years, and mean size was 7.5 mm (range, 1.6 to 16 mm). Presentation included palpable mass in 6 (22%), incidental in 6 (22%), and an imaging abnormality in 15 (56%) cases. Imaging impression included hematoma (24%), lymph node (10%), fat necrosis (10%), tortuous vessel (5%), and not provided in 52%. The lesions were classified on the basis of BCBx or BCBx and excision (available in 16 pts) as follows: 1 low-grade angiosarcoma, 8 angiolipomas, 6 capillary hemangiomas, 4 cavernous hemangiomas, 2 hemangiomas (not otherwise specified), 1 papillary endothelial hyperplasia, and 5 perilobular hemangiomas. The angiosarcoma was 9 mm, detected incidentally by magnetic resonance imaging, and showed dissection of stromal collagen, infiltration of glands, high cellularity, moderate cytologic atypia, scant mitotic activity, and Ki-67 reactivity of 10%. Among the 26 benign VLs, worrisome histologic features were noted in 14 on BCBx, including anastomosing vascular channels in 9, moderate cytologic atypia in 4, high cellularity in 2, Ki-67>10% in 2, mitotic activity in 1, and infiltration of glands in 1. Of the 12 VLs without worrisome features, the lesion extended to edge of core in 8, precluding complete evaluation. BCBx of VLs presents diagnostic challenges due to overlapping clinicopathologic and radiologic features with low-grade angiosarcoma. If completeness of removal is documented on BCBx, and cytoarchitectural changes are not worrisome, follow-up could be considered rather than excision. However, only 4 of these cases fulfilled those criteria.

AIMS/OBJECTIVE:To identify variables that can predict upgrade for magnetic resonance imaging (MRI)-detected atypical ductal hyperplasia (ADH). METHODS AND RESULTS/RESULTS:We reviewed 1655 MRI-guided core biopsies between 2005 and 2013, yielding 100 (6%) cases with ADH. The pathological features of ADH and MRI findings were recorded. An upgrade was considered when the subsequent surgical excision yielded invasive carcinoma (IC) or ductal carcinoma in situ (DCIS). The rate of ADH between institutions was 3.3-7.1%, with an average of 6%. A total of 15 (15%) cases had upgrade, 12 DCIS and three IC. When all cases were included, only increased number of involved cores was statistically significant (P = 0.02). When cases with concurrent lobular neoplasia (LN) were excluded (n = 14), increased number of ADH foci and increased number of involved cores were statistically significant (P = 0.002, P = 0.009). We analysed the data separately from a single institution (n = 61). Increased number of foci, increased number of total cores and involved cores and larger ADH size predicted upgrade with statistical significance. CONCLUSIONS:The incidence of ADH in MRI-guided core biopsy is rare. The rate of upgrade is comparable to mammographically detected ADH, warranting surgical excision. Similar to mammographically detected lesions, the volume of the ADH predicts the upgrade.

Sonography is useful in the evaluation of axillary lymph nodes in patients with breast cancer. In this pictorial essay, we review the range of grayscale and Doppler appearances of abnormal axillary lymph nodes on 2-dimensional and 3-dimensional imaging.

There are certain criteria to recommend surgical excision for lobular neoplasia diagnosed in mammographically detected core biopsy. The aims of this study are to explore the rate of upgrade of lobular neoplasia detected in magnetic resonance imaging (MRI)-guided biopsy and to investigate the clinicopathological and radiological features that could predict upgrade. We reviewed 1655 MRI-guided core biopsies yielding 63 (4%) cases of lobular neoplasia. Key clinical features were recorded. MRI findings including mass vs non-mass enhancement and the reason for biopsy were also recorded. An upgrade was defined as the presence of invasive carcinoma or ductal carcinoma in situ in subsequent surgical excision. The overall rate of lobular neoplasia in MRI-guided core biopsy ranged from 2 to 7%, with an average of 4%. A total of 15 (24%) cases had an upgrade, including 5 cases of invasive carcinoma and 10 cases of ductal carcinoma in situ. Pure lobular neoplasia was identified in 34 cases, 11 (32%) of which had upgrade. In this group, an ipsilateral concurrent or past history of breast cancer was found to be associated with a higher risk of upgrade (6/11, 55%) than contralateral breast cancer (1 of 12, 8%; P=0.03). To our knowledge, this is the largest series of lobular neoplasia diagnosed in MRI-guided core biopsy. The incidence of lobular neoplasia is relatively low. Lobular neoplasia detected in MRI-guided biopsy carries a high risk for upgrade warranting surgical excision. However, more cases from different types of institutions are needed to verify our results.

Retrospective analysis of 99 male autosomal dominant polycystic kidney disease (ADPKD) patients compared to an age-matched control population showed seminal vesicle ectasia >10 mm (megavesicle) in 23% (23/99) of ADPKD patients that was not present in any controls (P<.0001). Median (range) seminal vesicle convoluted tubule diameter in ADPKD patients was 4.2 (1.7-30) mm compared to 3.1 (1.7-6.8) mm in controls (P<.0001). Discrete cysts were identified in four ADPKD patients but in none of the control population (P=.12). Seminal megavesicles may explain the infertility sometimes observed in male ADPKD patients.