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.

BACKGROUND AND PURPOSE: Endoluminal reconstruction with flow-diverting stents represents a widely accepted technique for the treatment of complex intracranial aneurysms. This European registry study analyzed the initial experience of 15 neurovascular centers with the Flow-Redirection Intraluminal Device (FRED) system. MATERIALS AND METHODS: Consecutive patients with intracranial aneurysms treated with the FRED between February 2012 and March 2015 were retrospectively reviewed. Complications and adverse events, transient and permanent morbidity, mortality, and occlusion rates were evaluated. RESULTS: During the defined study period, 579 aneurysms in 531 patients (median age, 54 years; range, 13–86 years) were treated with the FRED. Seven percent of patients were treated in the acute phase (≤3 days) of aneurysm rupture. The median aneurysm size was 7.6 mm (range, 1–36.6 mm), and the median neck size 4.5 mm (range, 1–30 mm). Angiographic follow-up of >3 months was available for 516 (89.1%) aneurysms. There was progressive occlusion witnessed with time, with complete occlusion in 18 (20%) aneurysms followed for up to 90 ± 14 days, 141 (82.5%) for 180 ± 20 days, 116 (91.3%) for 1 year ± 24 days, and 122 (95.3%) aneurysms followed for >1 year. Transient and permanent morbidity occurred in 3.2% and 0.8% of procedures, respectively. The overall mortality rate was 1.5%. CONCLUSIONS: This retrospective study in real-world patients demonstrated the safety and efficacy of the FRED for the treatment of intracranial aneurysms. In most cases, treatment with a single FRED resulted in complete angiographic occlusion at 1 year.

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.

Objective To determine whether thrombectomy is safe in children up to 24 hours after onset of symptoms when selected by mismatch between clinical deficit and infarct. Methods A secondary analysis of the Save ChildS Study (January 2000–December 2018) was performed, including all pediatric patients (<18 years) diagnosed with arterial ischemic stroke who underwent endovascular recanalization at 27 European and United States stroke centers. Patients were included if they had a relevant mismatch between clinical deficit and infarct. Results Twenty children with a median age of 10.5 (interquartile range [IQR] 7–14.6) years were included. Of those, 7 were male (35%), and median time from onset to thrombectomy was 9.8 (IQR 7.8–16.2) hours. Neurologic outcome improved from a median Pediatric NIH Stroke Scale score of 12.0 (IQR 8.8–20.3) at admission to 2.0 (IQR 1.2–6.8) at day 7. Median modified Rankin Scale (mRS) score was 1.0 (IQR 0–1.6) at 3 months and 0.0 (IQR 0–1.0) at 24 months. One patient developed transient peri-interventional vasospasm; no other complications were observed. A comparison of the mRS score to the mRS score in the DAWN and DEFUSE 3 trials revealed a higher proportion of good outcomes in the pediatric compared to the adult study population. Conclusions Thrombectomy in pediatric ischemic stroke in an extended time window of up to 24 hours after onset of symptoms seems safe and neurologic outcomes are generally good if patients are selected by a mismatch between clinical deficit and infarct. Classification of Evidence This study provides Class IV evidence that for children with acute ischemic stroke with a mismatch between clinical deficit and infarct size, thrombectomy is safe.

Background and Purpose—: The recent Save ChildS study provides multicenter evidence for the use of mechanical thrombectomy in children with large vessel occlusion arterial ischemic stroke. However, device selection for thrombectomy may influence rates of recanalization, complications, and neurological outcomes, especially in pediatric patients of different ages. We, therefore, performed additional analyses of the Save ChildS data to investigate a possible association of different thrombectomy techniques and devices with angiographic and clinical outcome parameters. Methods—: The Save ChildS cohort study (January 2000–December 2018) analyzed data from 27 European and United States stroke centers and included all pediatric patients (<18 years), diagnosed with arterial ischemic stroke who underwent endovascular recanalization. Patients were grouped into first-line contact aspiration (A Direct Aspiration First Pass Technique [ADAPT]) and non-ADAPT groups as well as different stent retriever size groups. Associations with baseline characteristics, recanalization rates (modified Treatment in Cerebral Infarction), complication rates, and neurological outcome parameters (Pediatric National Institutes of Health Stroke Scale after 24 hours and 7 days; modified Rankin Scale and Pediatric Stroke Outcome Measure at discharge, after 6 and 24 months) were investigated. Results—: Seventy-three patients with a median age of 11.3 years were included. Currently available stent retrievers were used in 59 patients (80.8%), of which 4×20 mm (width×length) was the most frequently chosen size (36 patients =61%). A first-line ADAPT approach was used in 7 patients (9.6%), and 7 patients (9.6%) were treated with first-generation thrombectomy devices. In this study, a first-line ADAPT approach was neither associated with the rate of successful recanalization (ADAPT 85.7% versus 87.5% No ADAPT) nor with the complication rate or the neurological outcome. Moreover, there were no associations of stent retriever sizes with rates of recanalization, complication rates, or outcome parameters. Conclusions—: Our study suggests that neurological outcomes are generally good regardless of any specific device selection and suggests that it is important to offer thrombectomy in eligible children regardless of technique or device selection. Registration—: URL: https://www.drks.de/ ; Unique identifier: DRKS00016528.

In this study, we assessed the prognostic relevance of temporal muscle thickness (TMT), likely reflecting patient’s frailty, in patients with primary central nervous system lymphoma (PCNSL). In 128 newly diagnosed PCNSL patients TMT was analyzed on cranial magnetic resonance images. Predefined sex-specific TMT cutoff values were used to categorize the patient cohort. Survival analyses, using a log-rank test as well as Cox models adjusted for further prognostic parameters, were performed. The risk of death was significantly increased for PCNSL patients with reduced muscle thickness (hazard ratio of 3.189, 95% CI: 2–097–4.848, p < 0.001). Importantly, the results confirmed that TMT could be used as an independent prognostic marker upon multivariate Cox modeling (hazard ratio of 2.504, 95% CI: 1.608–3.911, p < 0.001) adjusting for sex, age at time of diagnosis, deep brain involvement of the PCNSL lesions, Eastern Cooperative Oncology Group (ECOG) performance status, and methotrexate-based chemotherapy. A TMT value below the sex-related cutoff value at the time of diagnosis is an independent adverse marker in patients with PCNSL. Thus, our results suggest the systematic inclusion of TMT in further translational and clinical studies designed to help validate its role as a prognostic biomarker.