Federated learning enables big data for rare cancer boundary detection.

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.

The federated tumor segmentation (FeTS) tool: an open-source solution to further solid tumor research.

Objective.De-centralized data analysis becomes an increasingly preferred option in the healthcare domain, as it alleviates the need for sharing primary patient data across collaborating institutions. This highlights the need for consistent harmonized data curation, pre-processing, and identification of regions of interest based on uniform criteria.Approach.Towards this end, this manuscript describes theFederatedTumorSegmentation (FeTS) tool, in terms of software architecture and functionality.Main results.The primary aim of the FeTS tool is to facilitate this harmonized processing and the generation of gold standard reference labels for tumor sub-compartments on brain magnetic resonance imaging, and further enable federated training of a tumor sub-compartment delineation model across numerous sites distributed across the globe, without the need to share patient data.Significance.Building upon existing open-source tools such as the Insight Toolkit and Qt, the FeTS tool is designed to enable training deep learning models targeting tumor delineation in either centralized or federated settings. The target audience of the FeTS tool is primarily the computational researcher interested in developing federated learning models, and interested in joining a global federation towards this effort. The tool is open sourced athttps://github.com/FETS-AI/Front-End.

Immune surveillance for six vaccinable pathogens using paired plasma and dried blood spots in HIV infected and uninfected children in Kinshasa.

Child vaccination reduces infant mortality rates. HIV-infected children present higher risk of diseases than non-infected. We report the protection coverage rates for 6 vaccine-preventable diseases in a paediatric population from the Democratic Republic of the Congo (DRC) and the impact of HIV infection, providing the first data on the validity of dried blood samples (DBS) to monitor the immune protection. During 2016-2018 DBS from 143 children/adolescents were collected in Kinshasa (DRC), being 52 HIV-infected. Forty-two had a paired plasma sample. Protective IgG was quantified (VirClia-IgG,VIRCELL) to obtain the optimal cut-off in IgG detection in DBS. ROC curves were generated with R software and statistical analyses with Stata. Protective IgG levels varied across pathogens, not reaching herd immunity. HIV-infected presented lower vaccine protection than uninfected for all analyzed pathogens, except rubella, with statistically significant differences for measles (30.8% vs. 53.8%; p = 0.008) and tetanus (3.8% vs. 22%; p = 0.0034). New cut-offs were calculated when using DBS to improve test performance. We reinforce the necessity to increase pediatric vaccination coverage in Kinshasa, especially in HIV seropositive, with less capacity to maintain adequate antibody levels. DBS were useful to monitor vaccination coverage in seroprevalence studies in resource-limited settings, after optimizing the cut-off value for each pathogen.

Predictors of retention in the prospective HIV prevention OKAPI cohort in Kinshasa.

Retention is a key element in HIV prevention programs. In Sub-Saharan Africa most data on retention come from HIV clinical trials or people living with HIV attending HIV treatment and control programs. Data from observational cohorts are less frequent. Retention at 6-/12-month follow-up and its predictors were analyzed in OKAPI prospective cohort. From April 2016 to April 2018, 797 participants aged 15-59 years attending HIV Voluntary Counseling and Testing in Kinshasa were interviewed about HIV-related knowledge and behaviors at baseline and at 6- and 12-month follow-ups. Retention rates were 57% and 27% at 6- and 12-month follow up; 22% of participants attended both visits. Retention at 6-month was significantly associated with 12-month retention. Retention was associated with low economic status, being studying, daily/weekly Internet access, previous HIV tests and aiming to share HIV test with partner. Contrarily, perceiving a good health, living far from an antiretroviral center, daily/weekly alcohol consumption and perceiving frequent HIV information were inversely associated with retention. In conclusion, a high attrition was found among people attending HIV testing participating in a prospective cohort in Kinshasa. Considering the low retention rates and the predictors found in this study, more HIV cohort studies in Kinshasa need to be evaluated to identify local factors and strategies that could improve retention if needed.

Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data.

Several studies underscore the potential of deep learning in identifying complex patterns, leading to diagnostic and prognostic biomarkers. Identifying sufficiently large and diverse datasets, required for training, is a significant challenge in medicine and can rarely be found in individual institutions. Multi-institutional collaborations based on centrally-shared patient data face privacy and ownership challenges. Federated learning is a novel paradigm for data-private multi-institutional collaborations, where model-learning leverages all available data without sharing data between institutions, by distributing the model-training to the data-owners and aggregating their results. We show that federated learning among 10 institutions results in models reaching 99% of the model quality achieved with centralized data, and evaluate generalizability on data from institutions outside the federation. We further investigate the effects of data distribution across collaborating institutions on model quality and learning patterns, indicating that increased access to data through data private multi-institutional collaborations can benefit model quality more than the errors introduced by the collaborative method. Finally, we compare with other collaborative-learning approaches demonstrating the superiority of federated learning, and discuss practical implementation considerations. Clinical adoption of federated learning is expected to lead to models trained on datasets of unprecedented size, hence have a catalytic impact towards precision/personalized medicine.

Systematic Evaluation of Image Tiling Adverse Effects on Deep Learning Semantic Segmentation.

Convolutional neural network (CNN) models obtain state of the art performance on image classification, localization, and segmentation tasks. Limitations in computer hardware, most notably memory size in deep learning accelerator cards, prevent relatively large images, such as those from medical and satellite imaging, from being processed as a whole in their original resolution. A fully convolutional topology, such as U-Net, is typically trained on down-sampled images and inferred on images of their original size and resolution, by simply dividing the larger image into smaller (typically overlapping) tiles, making predictions on these tiles, and stitching them back together as the prediction for the whole image. In this study, we show that this tiling technique combined with translationally-invariant nature of CNNs causes small, but relevant differences during inference that can be detrimental in the performance of the model. Here we quantify these variations in both medical (i.e., BraTS) and non-medical (i.e., satellite) images and show that training a 2D U-Net model on the whole image substantially improves the overall model performance. Finally, we compare 2D and 3D semantic segmentation models to show that providing CNN models with a wider context of the image in all three dimensions leads to more accurate and consistent predictions. Our results suggest that tiling the input to CNN models-while perhaps necessary to overcome the memory limitations in computer hardware-may lead to undesirable and unpredictable errors in the model's output that can only be adequately mitigated by increasing the input of the model to the largest possible tile size.

Multi-Institutional Deep Learning Modeling Without Sharing Patient Data: A Feasibility Study on Brain Tumor Segmentation.

Deep learning models for semantic segmentation of images require large amounts of data. In the medical imaging domain, acquiring sufficient data is a significant challenge. Labeling medical image data requires expert knowledge. Collaboration between institutions could address this challenge, but sharing medical data to a centralized location faces various legal, privacy, technical, and data-ownership challenges, especially among international institutions. In this study, we introduce the first use of federated learning for multi-institutional collaboration, enabling deep learning modeling without sharing patient data. Our quantitative results demonstrate that the performance of federated semantic segmentation models (Dice=0.852) on multimodal brain scans is similar to that of models trained by sharing data (Dice=0.862). We compare federated learning with two alternative collaborative learning methods and find that they fail to match the performance of federated learning.

Cómo avanzar en el reto del diagnóstico precoz de la infección por VIH / How to advance in meeting the challenge of early diagnosis of HIV infection

No disponible

Reduction in mortality associated with secondary cytomegalovirus prophylaxis after solid organ transplantation.

BACKGROUND: Cytomegalovirus (CMV) is the most important viral pathogen in solid organ transplant (SOT) recipients. The role of secondary CMV prophylaxis in this population remains unclear. METHODS: Retrospective cohort study in a single center. SOT recipients treated for CMV infection from 2007 to 2014 were studied to determine the efficacy and safety of secondary prophylaxis and its impact on graft loss and mortality. The outcome variable was CMV replication in the first 3 months after the end of therapy. Secondary variables were crude mortality and graft lost censored at 5 years after transplantation. Propensity score for the use of secondary prophylaxis was used to control selection bias. RESULTS: Of the 126 treated patients, 103 (83.1%) received CMV secondary prophylaxis. CMV relapse occurred in 44 (35.5%) patients. The use of secondary prophylaxis was not associated with fewer relapses (34.0% in patients with prophylaxis vs 42.9% in those without prophylaxis, P = .29). After a mean follow-up of 32.1 months, graft loss was not different between both groups but patient mortality was significantly lower in patients who received secondary prophylaxis (5.8% vs 28.6%, P = .003). CONCLUSION: Secondary prophylaxis did not prevent CMV infection relapse but it was associated with improved patient survival.

Prevalence and distribution of hepatitis C virus genotypes in Spain during the 2000-2015 period (the GEHEP 005 study).

We report the largest study on the prevalence and distribution of HCV genotypes in Spain (2000-2015), and we relate them with clinical, epidemiological and virological factors. Patients from 29 hospitals in 10 autonomous communities (Andalusia, Aragon, Castilla-Leon, Catalonia, Galicia, Canary Islands, Madrid Community, Valencian Community, Murcia Region and Basque Country) have been studied. Annual distribution of HCV genotypes and subtypes, as well as gender, age, transmission route, HIV and/or HBV coinfection, and treatment details were recorded. We included 48595 chronically HCV-infected patients with the following characteristics: median age 51 years (IQR, 44-58), 67.9% male, 19.1% HIV-coinfected, 23.5% HBV-coinfected. Parenteral transmission route was the most frequent (58.7%). Genotype distribution was 66.9% GT1 (24.9% subtype 1a and 37.9% subtype 1b), 2.8% GT2, 17.3% GT3, 11.4% GT4 and 0.1% GT5 and 0.02% GT6. LiPA was the most widely HCV genotyping test used (52.4%). HCV subtype 1a and genotypes 3 and 4 were closely associated with male gender, parenteral route of infection and HIV and HBV coinfection; in contrast, subtype 1b and genotype 2 were associated with female gender, nonparenteral route and mono-infection. Age was related to genotype distribution, and different patterns of distribution and biodiversity index were observed between different geographical areas. Finally, we describe how treatment and changes in transmission routes may have affected HCV genotype prevalence and distribution patterns. We present the most recent data on molecular epidemiology of hepatitis C virus in Spain. This study confirms that genotype distributions vary with age, sex, HIV and HBV coinfection and within geographical areas and epidemiological groups.

Rhodamine/Nanodiamond as a System Model for Drug Carrier.

In this paper we present some strategies that are being developed in our labs towards enabling nanodiamond-based applications for drug delivery. Rhodamine B (RhB) has been choosen as model molecule to study the loading of nanodiamonds with active moieties and the conditions for their controlled release. In order to test the chemical/physical interactions between functionalized detonation nanodiamond (DND) and complex molecules, we prepared and tested different RhB@DND systems, with RhB adsorbed or linked by ionic bonding to the DND surface. The chemical state of the DND surfaces before conjugation with the RhB molecules, and the chemical features of the DND-RhB interactions have been deeply analysed by coupling DND with Au nanoparticles and taking advantage of surface enhanced Raman spectroscopy SERS. The effects due to temperature and pH variations on the process of RhB release from the DND carrier have been also investigated. The amounts of released molecules are consistent with those required for effective drug action in conventional therapeutic applications, and this makes the DND promising nanostructured cargos for drug delivery applications.

Vaccine effectiveness in preventing laboratory-confirmed influenza in Navarre, Spain: 2013/14 mid-season analysis.

We estimate mid-2013/14 season vaccine effectiveness (VE) of the influenza trivalent vaccine in Navarre, Spain. Influenza-like illness cases attended in hospital (n=431) and primary healthcare (n=344) were included. The overall adjusted VE in preventing laboratory-confirmed influenza was 24% (95% CI: -14 to 50). The VE was 40% (95% CI: -12 to 68) against influenza A(H1)pdm09 and 13% (95% CI: -36 to 45) against influenza A(H3). These results suggest a moderate preventive effect against influenza A(H1)pdm09 and low protection against influenza A(H3).

Early estimates of influenza vaccine effectiveness in Navarre, Spain: 2012/13 mid-season analysis.

We present estimates of influenza vaccine effectiveness (VE) in Navarre, Spain, in the early 2012/13 season, which was dominated by influenza B. In a population-based cohort using electronic records from physicians, the adjusted VE in preventing influenzalike illness was 32% (95% confidence interval (CI): 15 to 46). In a nested test-negative case-control analysis the adjusted VE in preventing laboratory-confirmed influenza was 86% (95% CI: 45 to 96). These results suggest a high protective effect of the vaccine.

Decline in influenza vaccine effectiveness with time after vaccination, Navarre, Spain, season 2011/12.

This study evaluates the influenza vaccine effectiveness (VE) in preventing laboratory-confirmed cases in Navarre, Spain, in the 2011/12 season in which the peak was delayed until week 7 of 2012. We conducted a test-negative case­control study. Patients with influenza-like illness in hospitals and primary healthcare were swabbed for testing by reverse transcription-polymerase chain reaction. Influenza vaccination status and other covariates were obtained from healthcare databases. The vaccination status of confirmed cases and negative controls was compared after adjusting for potential confounders. VE was calculated as (1-odds ratio)x100. The 411 confirmed cases (93% influenza A(H3)) were compared with 346 controls. Most characterised viruses did not match the vaccine strains. The adjusted estimate of VE was 31% (95% confidence interval (CI): -21 to 60) for all patients, 44% (95% CI: -11 to 72) for those younger than 65 years and 19% (95% CI: -146 to 73) for those 65 or older. The VE was 61% (95% CI: 5 to 84) in the first 100 days after vaccination, 42% (95% CI: -39 to 75) between 100 and 119 days, and zero thereafter. This decline mainly affected people aged 65 or over. These results suggest a low preventive effect of the 2011/12 seasonal influenza vaccine, and a decline in VE with time since vaccination.

Telementoring in robotic surgery.

PURPOSE OF REVIEW: To provide an overview of the current concepts regarding telementoring with robotic surgery highlighting recent advances with respect to urological minimally invasive surgery (MIS). RECENT FINDINGS: As robotic surgery continues to evolve, telementoring will become a viable alternative to traditional on-site surgical proctoring. SUMMARY: MIS represents one of the most important breakthroughs in medicine over the past few decades. Newcomers to MIS need the guidance of more experienced, 'high volume' mentors to achieve the superior outcomes promised by MIS over conventional techniques.Telementoring, a subset of telemedicine, allows a surgeon at a remote site to offer intraoperative guidance via telecommunication networks. MIS lends itself well to telementoring techniques for several reasons; the primary surgeon performing MIS is working off of video images of the surgical field or images sent to a console. As such, the mentor is seeing the exact same images as the primary surgeon. In this review, we highlight many of the latest technologies in telemedicine, which are applicable to MIS and provide an overview of the pitfalls, which need to be overcome to make telementoring (and eventually telesurgery) a standard tool in the MIS arsenal.

Influenza outbreaks in nursing homes with high vaccination coverage in Navarre, Spain, 2011/12.

In the 2011/12 season, three influenza outbreaks were studied in nursing homes with high vaccination coverage in Navarre, Spain. Attack rates ranged from 2.9% to 67%. Influenza A/Stockholm/18/2011(H3N2) virus strain was isolated from the three outbreaks. Vaccination should be complemented with other hygiene measures in nursing homes. Early detection of influenza outbreaks in nursing homes can aid in their control.

Visualization of Electrostatic Dipoles in Molecular Dynamics of Metal Oxides.

Metal oxides are important for many technical applications. For example alumina (aluminum oxide) is the most commonly-used ceramic in microelectronic devices thanks to its excellent properties. Experimental studies of these materials are increasingly supplemented with computer simulations. Molecular dynamics (MD) simulations can reproduce the material behavior very well and are now reaching time scales relevant for interesting processes like crack propagation. In this work we focus on the visualization of induced electric dipole moments on oxygen atoms in crack propagation simulations. The straightforward visualization using glyphs for the individual atoms, simple shapes like spheres or arrows, is insufficient for providing information about the data set as a whole. As our contribution we show for the first time that fractional anisotropy values computed from the local neighborhood of individual atoms of MD simulation data depict important information about relevant properties of the field of induced electric dipole moments. Iso surfaces in the field of fractional anisotropy as well as adjustments of the glyph representation allow the user to identify regions of correlated orientation. We present novel and relevant findings for the application domain resulting from these visualizations, like the influence of mechanical forces on the electrostatic properties.

An increased burden of common and rare lipid-associated risk alleles contributes to the phenotypic spectrum of hypertriglyceridemia.

OBJECTIVE: Earlier studies have suggested that a common genetic architecture underlies the clinically heterogeneous polygenic Fredrickson hyperlipoproteinemia (HLP) phenotypes defined by hypertriglyceridemia (HTG). Here, we comprehensively analyzed 504 HLP-HTG patients and 1213 normotriglyceridemic controls and confirmed that a spectrum of common and rare lipid-associated variants underlies this heterogeneity. METHODS AND RESULTS: First, we demonstrated that genetic determinants of plasma lipids and lipoproteins, including common variants associated with plasma triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C) from the Global Lipids Genetics Consortium were associated with multiple HLP-HTG phenotypes. Second, we demonstrated that weighted risk scores composed of common TG-associated variants were distinctly increased across all HLP-HTG phenotypes compared with controls; weighted HDL-C and LDL-C risk scores were also increased, although to a less pronounced degree with some HLP-HTG phenotypes. Interestingly, decomposition of HDL-C and LDL-C risk scores revealed that pleiotropic variants (those jointly associated with TG) accounted for the greatest difference in HDL-C and LDL-C risk scores. The APOE E2/E2 genotype was significantly overrepresented in HLP type 3 versus other phenotypes. Finally, rare variants in 4 genes accumulated equally across HLP-HTG phenotypes. CONCLUSIONS: HTG susceptibility and phenotypic heterogeneity are both influenced by accumulation of common and rare TG-associated variants.