The correct interpretation of breast density is important in the assessment
of breast cancer risk. AI has been shown capable of accurately predicting
breast density, however, due to the differences in imaging characteristics
across mammography systems, models built using data from one system do not
generalize well to other systems. Though federated learning (FL) has emerged as
a way to improve the generalizability of AI without the need to share data, the
best way to preserve features from all training data during FL is an active
area of research. To explore FL methodology, the breast density classification
FL challenge was hosted in partnership with the American College of Radiology,
Harvard Medical School's Mass General Brigham, University of Colorado, NVIDIA,
and the National Institutes of Health National Cancer Institute. Challenge
participants were able to submit docker containers capable of implementing FL
on three simulated medical facilities, each containing a unique large
mammography dataset. The breast density FL challenge ran from June 15 to
September 5, 2022, attracting seven finalists from around the world. The
winning FL submission reached a linear kappa score of 0.653 on the challenge
test data and 0.413 on an external testing dataset, scoring comparably to a
model trained on the same data in a central location.

乳房密度的正确解释对于评估乳腺癌风险很重要。利用人工智能准确预测乳房密度的能力已经得到证明，然而，由于乳房 X 光成像系统的特性差异，使用一个系统的数据构建的模型在其他系统上泛化能力不强。联邦学习已被提出作为一种提高人工智能泛化性能的方法，而无需共享数据，但在联邦学习中如何保留所有训练数据的特征仍然是一个活跃的研究领域。为了探索联邦学习方法，与美国放射学学会，哈佛医学院的 Mass General Brigham，科罗拉多大学，NVIDIA 和国立卫生研究院国家癌症研究所合作举办了乳房密度分类联邦学习挑战赛。挑战参与者可以提交能够在三个模拟的医疗机构上实现联邦学习的 docker 容器，每个机构都包含一个独特的大型乳房 X 光数据集。乳房密度联邦学习挑战赛从 2022 年 6 月 15 日持续到 9 月 5 日，吸引了来自世界各地的七个入围者。获胜的联邦学习解决方案在挑战测试数据上达到了线性 kappa 分数 0.653，并在外部测试数据集上达到了 0.413 的分数，与在一个中心位置训练的模型相当。

2022 ACR-NCI-NVIDIA 联邦学习挑战中自动乳腺密度分类的公平评估结果

Fair Evaluation of Federated Learning Algorithms for Automated Breast  Density Classification: The Results of the 2022 ACR-NCI-NVIDIA Federated  Learning Challenge

Assessing breast cancer risk from imaging remains a subjective process, in
which radiologists employ simple computer aided detection (CAD) systems or
qualitative visual assessment to estimate breast percent density (PD). Machine
learning (ML) models have become the most promising way to quantify breast
cancer risk for early, accurate, and equitable diagnoses, but training such
models in medical research is often restricted to small, single-institution
data. Since patient demographics and imaging characteristics may vary
considerably across imaging sites, models trained on single-institution data
tend not to generalize well. In response to this problem, MammoDL is proposed,
an open-source software tool that leverages a U-Net architecture to accurately
estimate breast PD and complexity from mammography. With the Open Federated
Learning (OpenFL) library, this solution enables secure training on datasets
across multiple institutions. MammoDL is a leaner, more flexible model than its
predecessors, boasting improved generalization due to federation-enabled
training on larger, more representative datasets.

MammoDL 是一种开源软件工具，采用 U-Net 架构来准确估计乳腺密度和复杂性，通过 Open Federated Learning 在多个机构的数据集上实现了安全训练，与其他模型相比，该模型更轻便、更灵活，因为联邦学习使其在更大、更具代表性的数据集上训练，从而提高了泛化能力。