5 December 2022 in Nature Communications
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.
8 July 2025 in Nature Communications
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.
14 February 2015 in Insights into Imaging
The aim of this retrospective study was twofold: (1) to show the role of multi-detector computed tomography angiography (MDCTA) in the evaluation of intracranial arterial anatomic variations; (2) to highlight their clinical importance with illustrated example cases.
1 October 2009 in Head & Face Medicine
True ossification of the auricle with cartilage replacement by bone, is a very rare clinical entity and can result in an entirely rigid auricle.
25 November 2009 in Cases Journal
Introduction: Gorlin-Goltz syndrome is a rare hereditary disease. Pathogenesis of the syndrome is attributed to abnormalities in the long arm of chromosome 9 (q22.3-q31) and loss or mutations of human patched gene (PTCH1 gene). Multiple basal cell carcinomas (BCCs), odontogenic keratocysts, skeletal abnormalities, hyperkeratosis of palms and soles, intracranial ectopic calcifications of the falx cerebri and facial dysmorphism are considered the main clinical features. Diagnosis is based upon established major and minor clinical and radiological criteria and ideally confirmed by DNA analysis. Because of the different systems affected, a multidisciplinary approach team of various experts is required for a successful management. Case presentation: We report the case of a 19 year-old female who was involved in a car accident and found to present imaging findings of Gorlin-Goltz syndrome during a routine whole body computed tomography (CT) scan in order to exclude traumatic injuries. Conclusion: Radiologic findings of the syndrome are easily identifiable on CT scans and may prompt to early verification of the disease, which is very important for regular follow-up and better survival rates from the co-existent diseases.
23 July 2021 in Scientific Reports
VPVasilios Panagiotopoulos
CKChristina Kalogeropoulou
+3
To assess and compare all current computed tomography angiography (CTA) scoring systems for the diagnostic workup of brain death (BD) to digital subtraction angiography (DSA) and clinical tests. Fifty-two patients with a clinical suspicion of BD underwent CTA and subsequently DSA. The diagnostic performance of all current CTA scoring systems was compared to that of DSA, in all patients with a suspicion of BD. A comparison to clinical tests was made only in DSA-positive for BD patients (n = 49), since in DSA-negative BD patients (n = 3) clinical tests were not performed. Further subgroup analysis was performed in relation to skull defects (SDs) stratification. Statistical analysis was conducted by applying statistics-contingency tables, Cochran’s-Q test and McNemar’s test. The CTA -10, and -7- and all 4-point scoring systems, showed overall sensitivities of 81,6%, 87.8% and 95.9% respectively and 100% specificity, when compared to DSA. In patients with a clinical verification of BD, the CTA -10 and -7-point scoring systems were significantly inferior to clinical tests (p = 0.004 and p = 0.031), while the 4-point scoring systems showed no such difference (p = 0.5). All 4-point scoring systems showed 100% sensitivity in patients with a minor SD or no SD. In patients with a major SD, all CTA scoring systems (− 10, − 7- and all 4-point) were less sensitive (62.5%, 62.5% and 75% respectively). The presence of a major SD was associated with an 8 × relative risk for false negative results in all 4-point scoring systems. CTA showed excellent diagnostic performance in patients with a suspicion of BD. The 4-point CTA scoring systems are the most sensitive for the diagnosis of BD, although in patients with a major SD patient, the role of CTA is ambiguous.