Although machine learning (ML) has shown promise across disciplines, out-of-sample generalizability is concerning. This is currently addressed by sharing multi-site data, but such centralization is challenging/infeasible to scale due to various limitations. Federated ML (FL) provides an alternative paradigm for accurate and generalizable ML, by only sharing numerical model updates. Here we present the largest FL study to-date, involving data from 71 sites across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, reporting the largest such dataset in the literature (n = 6, 314). We demonstrate a 33% delineation improvement for the surgically targetable tumor, and 23% for the complete tumor extent, over a publicly trained model. We anticipate our study to: 1) enable more healthcare studies informed by large diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further analyses for glioblastoma by releasing our consensus model, and 3) demonstrate the FL effectiveness at such scale and task-complexity as a paradigm shift for multi-site collaborations, alleviating the need for data-sharing.

Computational competitions are the standard for benchmarking medical image analysis algorithms, but they typically use small curated test datasets acquired at a few centers, leaving a gap to the reality of diverse multicentric patient data. To this end, the Federated Tumor Segmentation (FeTS) Challenge represents the paradigm for real-world algorithmic performance evaluation. The FeTS challenge is a competition to benchmark (i) federated learning aggregation algorithms and (ii) state-of-the-art segmentation algorithms, across multiple international sites. Weight aggregation and client selection techniques were compared using a multicentric brain tumor dataset in realistic federated learning simulations, yielding benefits for adaptive weight aggregation, and efficiency gains through client sampling. Quantitative performance evaluation of state-of-the-art segmentation algorithms on data distributed internationally across 32 institutions yielded good generalization on average, albeit the worst-case performance revealed data-specific modes of failure. Similar multi-site setups can help validate the real-world utility of healthcare AI algorithms in the future.

Primary hyperparathyroidism is a systemic endocrine disease that has significant effects on bone remodeling through the action of parathyroid hormone on the musculoskeletal system. These findings are important as they can aid in distinguishing primary hyperparathyroidism from other forms of metabolic bone diseases and inform physicians regarding disease severity and complications. This pictorial essay compiles bone-imaging features with the aim of improving the diagnosis of skeletal involvement of primary hyperthyroidism.

The localization of persistent or recurrent disease in reoperative patients with primary hyperparathyroidism presents challenges for radiologists and surgeons alike. In this article, we summarize the relevant imaging modalities, compare their accuracy in identifying reoperative disease, and outline their advantages and disadvantages. Accurate localization by preoperative imaging is a predictor of operative success, whereas negative or discordant preoperative imaging is a risk factor for operative failure. Ultrasound is a common first-line modality because it is inexpensive, accessible, and radiation-free. However, it is highly operator-dependent and less accurate in the reoperative setting than in the primary setting. Sestamibi scintigraphy is superior to ultrasound in localizing reoperative disease but requires radiation, prolonged imaging times, and reader experience for accurate interpretation. Like ultrasound, sestamibi scintigraphy is less accurate in the reoperative setting because reoperative patients can exhibit distorted anatomy, altered perfusion of remaining glands, and interference of radiotracer uptake. Meanwhile, four-dimensional computed tomography (4DCT) is superior to ultrasound and sestamibi scintigraphy in localizing reoperative disease but requires the use of radiation and intravenous contrast. Both 4DCT and magnetic resonance imaging (MRI) do not significantly differ in accuracy between unexplored and reoperative patients. However, MRI is more costly, inaccessible, and time-consuming than 4DCT and is inappropriate as a first-line modality. Hybrid imaging with positron emission tomography and computed tomography (PET/CT) may be a promising second-line modality in the reoperative setting, particularly when first-line modalities are discordant or inconclusive. Lastly, selective venous sampling should be reserved for challenging cases in which noninvasive modalities are negative or discordant. In the challenging population of reoperative patients with PHPT, a multimodality approach that utilizes the expertise of high-volume centers can accurately localize persistent or recurrent disease and enable curative parathyroidectomy.