Open-source Repository Research Articles

CRaDLe: Deep code retrieval based on semantic Dependency Learning

Code retrieval is a common practice for programmers to reuse existing code snippets in the open-source repositories. Given a user query (i.e., a natural language description), code retrieval aims at searching the most relevant ones from a set of code snippets. The main challenge of effective code retrieval lies in mitigating the semantic gap between natural language descriptions and code snippets. With the ever-increasing amount of available open-source code, recent studies resort to neural networks to learn the semantic matching relationships between the two sources. The statement-level dependency information, which highlights the dependency relations among the program statements during the execution, reflects the structural importance of one statement in the code, which is favorable for accurately capturing the code semantics but has never been explored for the code retrieval task. In this paper, we propose CRaDLe, a novel approach for Code Retrieval based on statement-level semantic Dependency Learning. Specifically, CRaDLe distills code representations through fusing both the dependency and semantic information at the statement level, and then learns a unified vector representation for each code and description pair for modeling the matching relationship. Comprehensive experiments and analysis on real-world datasets show that the proposed approach can accurately retrieve code snippets for a given query and significantly outperform the state-of-the-art approaches on the task.

From SIR to SEAIRD: A novel data-driven modeling approach based on the Grey-box System Theory to predict the dynamics of COVID-19.

Common compartmental modeling for COVID-19 is based on a priori knowledge and numerous assumptions. Additionally, they do not systematically incorporate asymptomatic cases. Our study aimed at providing a framework for data-driven approaches, by leveraging the strengths of the grey-box system theory or grey-box identification, known for its robustness in problem solving under partial, incomplete, or uncertain data. Empirical data on confirmed cases and deaths, extracted from an open source repository were used to develop the SEAIRD compartment model. Adjustments were made to fit current knowledge on the COVID-19 behavior. The model was implemented and solved using an Ordinary Differential Equation solver and an optimization tool. A cross-validation technique was applied, and the coefficient of determination R2 was computed in order to evaluate the goodness-of-fit of the model. Key epidemiological parameters were finally estimated and we provided the rationale for the construction of SEAIRD model. When applied to Brazil’s cases, SEAIRD produced an excellent agreement to the data, with an R2 ≥ 90%. The probability of COVID-19 transmission was generally high (≥ 95%). On the basis of a 20-day modeling data, the incidence rate of COVID-19 was as low as 3 infected cases per 100,000 exposed persons in Brazil and France. Within the same time frame, the fatality rate of COVID-19 was the highest in France (16.4%) followed by Brazil (6.9%), and the lowest in Russia (≤ 1%). SEAIRD represents an asset for modeling infectious diseases in their dynamical stable phase, especially for new viruses when pathophysiology knowledge is very limited.Supplementary InformationThe online version contains supplementary material available at 10.1007/s10489-021-02379-2.

An Anatomical Atlas-Based Scaling Study for Quantifying Muscle and Hip Joint Contact Forces in Above and Through-Knee Amputees Using Validated Musculoskeletal Modelling.

Customisation of musculoskeletal modelling using magnetic resonance imaging (MRI) significantly improves the model accuracy, but the process is time consuming and computationally intensive. This study hypothesizes that linear scaling to a lower limb amputee model with anthropometric similarity can accurately predict muscle and joint contact forces. An MRI-based anatomical atlas, comprising 18 trans-femoral and through-knee traumatic lower limb amputee models, is developed. Gait data, using a 10-camera motion capture system with two force plates, and surface electromyography (EMG) data were collected. Muscle and hip joint contact forces were quantified using musculoskeletal modelling. The predicted muscle activations from the subject-specific models were validated using EMG recordings. Anthropometry based multiple linear regression models, which minimize errors in force predictions, are presented. All predictions showed excellent (error interval c = 0-0.15), very good (c = 0.15-0.30) or good (c = 0.30-0.45) similarity to the EMG data, demonstrating accurate computation of muscle activations. The primary predictors of discrepancies in force predictions were differences in pelvis width (p < 0.001), body mass index (BMI, p < 0.001) and stump length to pelvis width ratio (p < 0.001) between the respective individual and underlying dataset. Linear scaling to a model with the most similar pelvis width, BMI and stump length to pelvis width ratio results in modelling outcomes with minimal errors. This study provides robust tools to perform accurate analyses of musculoskeletal mechanics for high-functioning lower limb military amputees, thus facilitating the further understanding and improvement of the amputee's function. The atlas is available in an open source repository.

Characteristics, potentials, and limitations of open-source Simulink projects for empirical research

Simulink is an example of a successful application of the paradigm of model-based development into industrial practice. Numerous companies create and maintain Simulink projects for modeling software-intensive embedded systems, aiming at early validation and automated code generation. However, Simulink projects are not as easily available as code-based ones, which profit from large publicly accessible open-source repositories, thus curbing empirical research. In this paper, we investigate a set of 1734 freely available Simulink models from 194 projects and analyze their suitability for empirical research. We analyze the projects considering (1) their development context, (2) their complexity in terms of size and organization within projects, and (3) their evolution over time. Our results show that there are both limitations and potentials for empirical research. On the one hand, some application domains dominate the development context, and there is a large number of models that can be considered toy examples of limited practical relevance. These often stem from an academic context, consist of only a few Simulink blocks, and are no longer (or have never been) under active development or maintenance. On the other hand, we found that a subset of the analyzed models is of considerable size and complexity. There are models comprising several thousands of blocks, some of them highly modularized by hierarchically organized Simulink subsystems. Likewise, some of the models expose an active maintenance span of several years, which indicates that they are used as primary development artifacts throughout a project’s lifecycle. According to a discussion of our results with a domain expert, many models can be considered mature enough for quality analysis purposes, and they expose characteristics that can be considered representative for industry-scale models. Thus, we are confident that a subset of the models is suitable for empirical research. More generally, using a publicly available model corpus or a dedicated subset enables researchers to replicate findings, publish subsequent studies, and use them for validation purposes. We publish our dataset for the sake of replicating our results and fostering future empirical research.

LINDA-BN: An interpretable probabilistic approach for demystifying black-box predictive models

The use of sophisticated machine learning models for critical decision-making faces the challenge that these models are often applied as a ‘black-box’. This has led to an increased interest in interpretable machine learning, where post-hoc model-agnostic algorithms present a useful mechanism for generating interpretations of complex learning models. This paper proposes a novel approach based on Bayesian Networks to generate local post-hoc model-agnostic interpretations of a black-box predictive model. Consequently, the proposed approach presents features that are conditionally dependent between each other and that are directly influencing the class variable. This enables the decision-maker to better understand how features are related and why a certain prediction was made. Compared to the existing post-hoc interpretation methods, the contribution of our approach is three-fold: (1) as a probabilistic graphical model, the extracted Bayesian network can provide interpretations through conditional dependencies in a graphical structure regarding what input features and how/why they contributed to a prediction; (2) for complex decision problems with many features, a Markov blanket can be generated from the extracted Bayesian network to provide interpretations with a focused view on those input features that directly contributed to a prediction; (3) the extracted Bayesian network enables the identification of four different rules which can inform the decision-maker about the confidence level in a prediction, thus helping the decision-maker assess the reliability of predictions learned by a black-box model. We implemented the proposed approach, applied it in the context of two well-known public datasets and analysed the results, which are made available in an open-source repository: https://github.com/catarina-moreira/LINDA_DSS.

An empirical investigation of organic software product lines

Software product line engineering is a best practice for managing reuse in families of software systems that is increasingly being applied to novel and emerging domains. In this work we investigate the use of software product line engineering in one of these new domains, synthetic biology. In synthetic biology living organisms are programmed to perform new functions or improve existing functions. These programs are designed and constructed using small building blocks made out of DNA. We conjecture that there are families of products that consist of common and variable DNA parts, and we can leverage product line engineering to help synthetic biologists build, evolve, and reuse DNA parts. In this paper we perform an investigation of domain engineering that leverages an open-source repository of more than 45,000 reusable DNA parts. We show the feasibility of these new types of product line models by identifying features and related artifacts in up to 93.5% of products, and that there is indeed both commonality and variability. We then construct feature models for four commonly engineered functions leading to product lines ranging from 10 to 7.5 × 1020 products. In a case study we demonstrate how we can use the feature models to help guide new experimentation in aspects of application engineering. Finally, in an empirical study we demonstrate the effectiveness and efficiency of automated reverse engineering on both complete and incomplete sets of products. In the process of these studies, we highlight key challenges and uncovered limitations of existing SPL techniques and tools which provide a roadmap for making SPL engineering applicable to new and emerging domains.

Spatial variation of physicochemical parameters in a constructed wetland for wastewater treatment: An example of the use of the R programming language

Introduction: The implementation of wastewater treatment systems such as constructed wetlands has a growing interest in the last decade due to its low cost and high effectiveness in treating industrial and residential wastewater. Objective: To evaluate the spatial variation of physicochemical parameters in a constructed wetland system of sub-superficial flow of Pennisetum alopecuroides (Pennisetum) and a Control (unplanted). The purpose is to provide an analysis of spatial dynamic of physicochemical parameters using R programming language. Methods: Each of the cells (Pennisetum and Control) had 12 piezometers, organized in three columns and four rows with a separation distance of 3,25m and 4,35m, respectively. The turbidity, biochemical oxygen demand (BOD), chemical oxygen demand (COD), total Kjeldahl nitrogen (TKN), ammoniacal nitrogen (N-NH4), organic nitrogen (N-org.) and phosphorous (P-PO4-3) were measured in water under in-flow and out-flow of both conditions Control and Pennisetum (n= 8). Additionally, the oxidation-reduction potential (ORP), dissolved oxygen (DO), conductivity, pH and water temperature, were measured (n= 167) in the piezometers. Results: No statistically significant differences between cells for TKN, N-NH4, conductivity, turbidity, BOD, and COD were found; but both Control and Pennisetum cells showed a significant reduction in these parameters (P&lt;0,05). Overall, TKN and N-NH4 removal were from 65,8 to 84,1% and 67,5 to 90,8%, respectively; and decrease in turbidity, conductivity, BOD, and COD, were between 95,1-95,4%; 15-22,4%; 65,2-77,9% and 57,4-60,3% respectively. Both cells showed ORP increasing gradient along the water-flow direction, contrary to conductivity (p&lt;0,05). However, OD, pH and temperature were inconsistent in the direction of the water flow in both cells. Conclusions: Pennisetum demonstrated pollutant removal efficiency, but presented results similar to the control cells, therefore, remains unclear if it is a superior option or not. Spatial variation analysis did not reflect any obstruction of flow along the CWs; but some preferential flow paths can be distinguished. An open-source repository of R was provided.

Evaluation of the Use of Single- and Multi-Magnification Convolutional Neural Networks for the Determination and Quantitation of Lesions in Nonclinical Pathology Studies.

Digital pathology platforms with integrated artificial intelligence have the potential to increase the efficiency of the nonclinical pathologist’s workflow through screening and prioritizing slides with lesions and highlighting areas with specific lesions for review. Herein, we describe the comparison of various single- and multi-magnification convolutional neural network (CNN) architectures to accelerate the detection of lesions in tissues. Different models were evaluated for defining performance characteristics and efficiency in accurately identifying lesions in 5 key rat organs (liver, kidney, heart, lung, and brain). Cohorts for liver and kidney were collected from TG-GATEs open-source repository, and heart, lung, and brain from internally selected R&D studies. Annotations were performed, and models were trained on each of the available lesion classes in the available organs. Various class-consolidation approaches were evaluated from generalized lesion detection to individual lesion detections. The relationship between the amount of annotated lesions and the precision/accuracy of model performance is elucidated. The utility of multi-magnification CNN implementations in specific tissue subtypes is also demonstrated. The use of these CNN-based models offers users the ability to apply generalized lesion detection to whole-slide images, with the potential to generate novel quantitative data that would not be possible with conventional image analysis techniques.

Evidence of Brood Parasitism and Quantification of Rangewide Overlap between the Olive Warbler and Brown-headed Cowbird

Brood parasitism is a fascinating natural history phenomenon that provides a window into the coevolution of antagonistic interactions. Many ecological and evolutionary aspects of brood parasitism remain unknown, and new hosts of brood parasites are still being discovered. We document the second instance of brood parasitism of the Olive Warbler (Peucedramus taeniatus) by the Brown-headed Cowbird (Molothrus ater). Apparent lack of habitat and elevation overlap across the ranges of these two species prompted us to examine how frequently they come into contact. We analyzed &gt;3 million Olive Warbler and Brown-headed Cowbird occurrence records from two open-source repositories, eBird and GBIF, to examine both synchronous and asynchronous locality overlap during the breeding season. We found that the two species were documented together simultaneously in only 3.1% of analyzed instances, but that they co-occurred in similar habitat types and/or at similar elevations at 11.8% of localities analyzed across the Olive Warbler’s range. Additional research on aspects of ecology and evolution, such as host selection, the cowbird’s diurnal patterns of movement, and the dynamics of intracellular pathogens infecting brood parasites and their hosts, may shed light more broadly on the ecological interactions and mechanisms underlying brood parasitism.

Automated Head Tissue Modelling Based on Structural Magnetic Resonance Images for Electroencephalographic Source Reconstruction

In the last years, technological advancements for the analysis of electroencephalography (EEG) recordings have permitted to investigate neural activity and connectivity in the human brain with unprecedented precision and reliability. A crucial element for accurate EEG source reconstruction is the construction of a realistic head model, incorporating information on electrode positions and head tissue distribution. In this paper, we introduce MR-TIM, a toolbox for head tissue modelling from structural magnetic resonance (MR) images. The toolbox consists of three modules: 1) image pre-processing – the raw MR image is denoised and prepared for further analyses; 2) tissue probability mapping – template tissue probability maps (TPMs) in individual space are generated from the MR image; 3) tissue segmentation – information from all the TPMs is integrated such that each voxel in the MR image is assigned to a specific tissue. MR-TIM generates highly realistic 3D masks, five of which are associated with brain structures (brain and cerebellar grey matter, brain and cerebellar white matter, and brainstem) and the remaining seven with other head tissues (cerebrospinal fluid, spongy and compact bones, eyes, muscle, fat and skin). Our validation, conducted on MR images collected in healthy volunteers and patients as well as an MR template image from an open-source repository, demonstrates that MR-TIM is more accurate than alternative approaches for whole-head tissue segmentation. We hope that MR-TIM, by yielding an increased precision in head modelling, will contribute to a more widespread use of EEG as a brain imaging technique.

Knowledge Network Embedding of Transcriptomic Data from Spaceflown Mice Uncovers Signs and Symptoms Associated with Terrestrial Diseases.

There has long been an interest in understanding how the hazards from spaceflight may trigger or exacerbate human diseases. With the goal of advancing our knowledge on physiological changes during space travel, NASA GeneLab provides an open-source repository of multi-omics data from real and simulated spaceflight studies. Alone, this data enables identification of biological changes during spaceflight, but cannot infer how that may impact an astronaut at the phenotypic level. To bridge this gap, Scalable Precision Medicine Oriented Knowledge Engine (SPOKE), a heterogeneous knowledge graph connecting biological and clinical data from over 30 databases, was used in combination with GeneLab transcriptomic data from six studies. This integration identified critical symptoms and physiological changes incurred during spaceflight.

SCHISTOX: An individual based model for the epidemiology and control of schistosomiasis.

A stochastic individual based model, SCHISTOX, has been developed for the study of schistosome transmission dynamics and the impact of control by mass drug administration. More novel aspects that can be investigated include individual level adherence and access to treatment, multiple communities, human sex population dynamics, and implementation of a potential vaccine. Many of the model parameters have been estimated within previous studies and have been shown to vary between communities, such as the age-specific contact rates governing the age profiles of infection. However, uncertainty remains as there are wide ranges for certain parameter values and a few remain relatively unknown. We analyse the model dynamics by parameterizing it with published parameter values. We also discuss the development of SCHISTOX in the form of a publicly available open-source GitHub repository. The next key development stage involves validating the model by calibrating to epidemiological data.

Automating Test Case Identification in Java Open Source Projects on GitHub

Software testing is one of the very important Quality Assurance (QA) components. A lot of researchers deal with the testing process in terms of tester motivation and how tests should or should not be written. However, it is not known from the recommendations how the tests are written in real projects. In this paper, the following was investigated: (i) the denotation of the word "test" in different natural languages; (ii) whether the number of occurrences of the word "test" correlates with the number of test cases; and (iii) what testing frameworks are mostly used. The analysis was performed on 38 GitHub open source repositories thoroughly selected from the set of 4.3M GitHub projects. We analyzed 20,340 test cases in 803 classes manually and 170k classes using an automated approach. The results show that: (i) there exists a weak correlation (r = 0.655) between the number of occurrences of the word "test" and the number of test cases in a class; (ii) the proposed algorithm using static file analysis correctly detected 97% of test cases; (iii) 15% of the analyzed classes used main() function whose represent regular Java programs that test the production code without using any third-party framework. The identification of such tests is very complex due to implementation diversity. The results may be leveraged to more quickly identify and locate test cases in a repository, to understand practices in customized testing solutions, and to mine tests to improve program comprehension in the future.

COVID-19: A forest fire rather than a wave?

As I write this piece, many countries around the world are being described as experiencing a «second wave» of the COVID-19 pandemic. For example, on 19 September 2020, the UK Prime Minister Boris Johnson said: «We are now seeing a second wave coming in. We’ve seen it in France, in Spain, across Europe. It’s been absolutely inevitable, I’m afraid, that we would see it in this country».&#x0D; Metaphors are crucial tools for communication and thinking, and can be particularly useful in public health communication. For example, the «second wave» metaphor suggests that there is renewed danger and threat from the virus, and may therefore encourage compliance with measures aimed at reducing transmission. However, all metaphors have both strengths and limitations, and the potential to be used both to enlighten and to obfuscate. The metaphor of the pandemic as a series of waves suggests that changes in the number of infections are due to the virus itself (cf. the idea that it may be seasonal), rather than the result of actions taken to slow its spread. In this sense, this metaphor is inaccurate. As Dr. Margaret Harris from the World Health Organization put it, «We are in the first wave. There is going to be one big wave». In addition, precisely because waves follow one another uncontrollably, this metaphor can be used strategically to present new increases in infection as inevitable, as in Boris Johnson’s statement, and thus to deflect responsibility from governments and their policies.&#x0D; As with any other complex and long-term problem, different metaphors are needed to capture different aspects of the pandemic, convey different messages, and address different audiences.&#x0D; Based on the analysis of two different datasets (the #ReframeCovid multilingual metaphor collection – an open-source repository of non-war-related language on COVID-19 – and the English Coronavirus Corpus – a multi-million-word database of news articles in English since January 2020 – I suggest that the metaphor of COVID-19 as a fire, and specifically a forest fire, is particularly apt and versatile. Forest fires are dangerous and hard to control. However, they can be controlled, with prompt and appropriate action. They can even be prevented, by looking after the land properly, protecting the environment, and educating citizens to behave responsibly.&#x0D; Indeed, forest fire metaphors for COVID-19 have been used since the start of the pandemic for multiple purposes, including to: convey danger and urgency (e.g., COVID-19 as a «forest fire that may not slow down»); distinguish between different phases of the pandemic (e.g., «a fire raging» vs. «embers» that must be stopped from causing a new fire); explain how contagion happens and the role of individuals within that (e.g., people as trees in a forest catching fire one after the other, or as breathing out «invisible embers»); justify measures for reducing contagion (e.g., social distancing as «fire lines» in a forest); connect the pandemic with health inequalities (e.g., pointing out that, like a fire, COVID-19 spreads more easily when people live in overcrowded conditions); and outline post-pandemic futures (e.g., when an Italian commentator pointed out that everyone has to contribute to the reclamation of the soil – bonifica del terreno – after the end of the pandemic, to prevent future ones).&#x0D; Of course, no metaphor is suitable for all purposes or all audiences. For example, the metaphor of people as trees in a forest fire does not easily account for asymptomatic transmission. And the use of forest fire metaphors may be inappropriate in parts of the world that have been dramatically affected by literal fires, such as some parts of Australia in 2019-2020. However, a well-informed and context-sensitive approach to metaphor selection can be an important and effective part of public health messaging.

Improvements and limitations in developing multivariate models of hemorrhage and transfusion risk for the obstetric population

Maternal hemorrhage protocols involve risk screening. These protocols prepare clinicians for potential hemorrhage and transfusion in individual patients. Patient-specific estimation and stratification of risk may improve maternal outcomes. Prediction models for hemorrhage and transfusion were trained and tested in a data set of 74 variables from 63 973 deliveries (97.6% of the source population of 65 560 deliveries included in a perinatal database from an academic urban delivery center) with sufficient data at pertinent time points: antepartum, peripartum, and postpartum. Hemorrhage and transfusion were present in 6% and 1.6% of deliveries, respectively. Model performance was evaluated with the receiver operating characteristic (ROC), precision-recall curves, and the Hosmer-Lemeshow calibration statistic. For hemorrhage risk prediction, logistic regression model discrimination showed ROCs of 0.633, 0.643, and 0.661 for the antepartum, peripartum, and postpartum models, respectively. These improve upon the California Maternal Quality Care Collaborative (CMQCC) accuracy of 0.613 for hemorrhage. Predictions of transfusion resulted in ROCs of 0.806, 0.822, and 0.854 for the antepartum, peripartum, and postpartum models, respectively. Previously described and new risk factors were identified. Models were not well calibrated with Hosmer-Lemeshow statistic P values between .001 and .6. Our models improve on existing risk assessment; however, further enhancement might require the inclusion of more granular, dynamic data. With the goal of increasing translatability, this work was distilled to an online open-source repository, including a form allowing risk factor inputs and outputs of CMQCC risk, alongside our numerical risk estimation and stratification of hemorrhage and transfusion.

Investigating the information value of different sources of evidence of developers’ expertise for bug assignment in open‐source projects

Bug assignment (BA), the process of ranking developers according to their potential ability to fix a given bug, is an important software-engineering task. BA usually requires the development of an expertise profile for each developer, and formulation of a similarity metric to estimate the relevance of developers to the bug. This needs us to answer the following question: ‘what is the information value of various contributions of developers in BA research?’ We address this question by making the following contributions. (i) We enhance the expertise metric of our prior work, vocabulary and time-based BA, to consider information regarding various sources of expertise with different importance. We show that this can improve the effectiveness of bug-assignment process. (ii) Using this ‘Multisource’ expertise metric, we investigate the information value of different pieces of information in open-source repositories for BA. We show that in addition to bug-fixing contributions, other technical and even social contributions of developers within the version-control system are useful information for BA. (iii) We provide a curated, up-to-date data set including technical information of 13 popular open-source projects in Github. To the best of our knowledge, this is the most comprehensive data set, currently available for bug-assignment research.

A spatiotemporal data collection of viral cases for COVID-19 rapid response

ABSTRACT Under the global health crisis of COVID-19, timely, and accurate epidemic data are important for observation, monitoring, analyzing, modeling, predicting, and mitigating impacts. Viral case data can be jointly analyzed with relevant factors for various applications in the context of the pandemic. Current COVID-19 case data are scattered across a variety of data sources which may consist of low data quality accompanied by inconsistent data structures. To address this shortcoming, a multi-scale spatiotemporal data product is proposed as a public repository platform, based on a spatiotemporal cube, and allows the integration of different data sources by adopting various data standards. Within the spatiotemporal cube, a comprehensive data processing workflow gathers disparate COVID-19 epidemic datasets at the global, national, provincial/state, county, and city levels. This proposed framework is supported by an automatic update with a 2-h frequency and the crowdsourcing validation team to produce and update data on a daily time step. This rapid-response dataset allows the integration of other relevant socio-economic and environmental factors for spatiotemporal analysis. The data is available in Harvard Dataverse platform (https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/8HGECN) and GitHub open source repository (https://github.com/stccenter/COVID-19-Data).

High-Resolution Ultrasound Imaging Enabled by Random Interference and Joint Image Reconstruction.

In ultrasound, wave interference is an undesirable effect that degrades the resolution of the images. We have recently shown that a wavefront of random interference can be used to reconstruct high-resolution ultrasound images. In this study, we further improve the resolution of interference-based ultrasound imaging by proposing a joint image reconstruction scheme. The proposed reconstruction scheme utilizes radio frequency (RF) signals from all elements of the sensor array in a joint optimization problem to directly reconstruct the final high-resolution image. By jointly processing array signals, we significantly improved the resolution of interference-based imaging. We compare the proposed joint reconstruction method with popular beamforming techniques and the previously proposed interference-based compound method. The simulation study suggests that, among the different reconstruction methods, the joint reconstruction method has the lowest mean-squared error (MSE), the best peak signal-to-noise ratio (PSNR), and the best signal-to-noise ratio (SNR). Similarly, the joint reconstruction method has an exceptional structural similarity index (SSIM) of 0.998. Experimental studies showed that the quality of images significantly improved when compared to other image reconstruction methods. Furthermore, we share our simulation codes as an open-source repository in support of reproducible research.

Intraprostatic Tumor Segmentation on PSMA PET Images in Patients with Primary Prostate Cancer with a Convolutional Neural Network.

Accurate delineation of the intraprostatic gross tumor volume (GTV) is a prerequisite for treatment approaches in patients with primary prostate cancer (PCa). Prostate-specific membrane antigen PET (PSMA PET) may outperform MRI in GTV detection. However, visual GTV delineation underlies interobserver heterogeneity and is time consuming. The aim of this study was to develop a convolutional neural network (CNN) for automated segmentation of intraprostatic tumor (GTV-CNN) in PSMA PET. Methods: The CNN (3D U-Net) was trained on the 68Ga-PSMA PET images of 152 patients from 2 different institutions, and the training labels were generated manually using a validated technique. The CNN was tested on 2 independent internal (cohort 1: 68Ga-PSMA PET, n = 18 and cohort 2: 18F-PSMA PET, n = 19) and 1 external (cohort 3: 68Ga-PSMA PET, n = 20) test datasets. Accordance between manual contours and GTV-CNN was assessed with the Dice-Sørensen coefficient (DSC). Sensitivity and specificity were calculated for the 2 internal test datasets (cohort 1: n = 18, cohort 2: n = 11) using whole-mount histology. Results: The median DSCs for cohorts 1-3 were 0.84 (range: 0.32-0.95), 0.81 (range: 0.28-0.93), and 0.83 (range: 0.32-0.93), respectively. Sensitivities and specificities for the GTV-CNN were comparable with manual expert contours: 0.98 and 0.76 (cohort 1) and 1 and 0.57 (cohort 2), respectively. Computation time was around 6 s for a standard dataset. Conclusion: The application of a CNN for automated contouring of intraprostatic GTV in 68Ga-PSMA and 18F-PSMA PET images resulted in a high concordance with expert contours and in high sensitivities and specificities in comparison with histology as a reference. This robust, accurate and fast technique may be implemented for treatment concepts in primary prostate cancer. The trained model and the study's source code are available in an open source repository.

Driving Green Open Access

Ubiquity Press was founded to catalyze and accelerate the adoption of open access. Until now we have followed the gold and diamond models, and currently support OA publication in over 700 journals, plus hundreds of book publications per year. We are now expanding our platform to fully support and drive the uptake and effectiveness of green open access also. In this presentation I will explain how we are doing this by integrating an open source repository solution with our platform. I will detail the features being added to especially enhance green OA, as part of the Advancing Hyku project with funding from Arcadia, a charitable fund of philanthropists Lisbet Rausing and Peter Baldwin, and in partnership with the University of Virginia and the British Library.