Breast Imaging Fellowship Training in the United States: A National Survey of Fellowship Program Directors.

OBJECTIVE: To provide an updated characterization of breast imaging fellowship programs in the United States to identify opportunities for improvement and standardization. METHODS: An anonymous survey was e-mailed to program directors of breast imaging fellowship programs listed on the Society of Breast Imaging website. The survey was open from April 23, 2021, through May 27, 2021. The survey was deemed exempt by the IRB. RESULTS: Forty-seven of 80 (59%) program directors responded, of which 36/47 (77%) represented programs dedicated 100% to breast imaging, and 11/47 (23%) represented programs dedicated 50%-75% to breast imaging. Common elements to most programs include tumor boards (47/47, 100%), journal clubs (39/47, 83%), case-based teaching sessions (35/47, 74%), didactic lectures (40/47, 85%), and participation in radiology-pathology conferences (29/47, 62%). Mammography Quality and Standards Act audit training (22/47, 47%), mammography quality control training (22/47, 47%), and formal communication training (19/47, 40%) were less common. Most programs provide exposure to wire (42/47, 89%) and wire-free localization procedures (45/47, 96%), but exposure to contrast-enhanced mammography (13/47, 28%) and molecular breast imaging (4/47, 9%) was limited. A small majority of programs (25/47, 53%) do not require weekday call; however, more (31/47, 66%) have weekend call responsibilities. Many programs (29/47, 62%) offer at least 3 weeks of elective time, which may be clinical or nonclinical. CONCLUSION: Breast imaging fellowship programs vary in curricula, modality exposure, and academic policies. The results of this survey can help guide further efforts to standardize and optimize fellowship training.

Deep Neural Networks Improve Radiologists' Performance in Breast Cancer Screening.

We present a deep convolutional neural network for breast cancer screening exam classification, trained, and evaluated on over 200000 exams (over 1000000 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. 1) 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. 2) 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. 3) Pretraining the network on screening BI-RADS classification, a related task with more noisy labels. 4) 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/breast_cancer_classifier.