Search

Methods for out-of-distribution (OOD) detection that scale to 3D data are crucial components of any real-world clinical deep learning system. Classic denoising diffusion probabilistic models (DDPMs) have been recently proposed as a robust way to perform reconstruction-based OOD detection on 2D datasets, but do not trivially scale to 3D data. In this work, we propose to use Latent Diffusion Models (LDMs), which enable the scaling of DDPMs to high-resolution 3D medical data. We validate the proposed approach on near- and far-OOD datasets and compare it to a recently proposed, 3D-enabled approach using Latent Transformer Models (LTMs). Not only does the proposed LDM-based approach achieve statistically significant better performance, it also shows less sensitivity to the underlying latent representation, more favourable memory scaling, and produces better spatial anomaly maps. Code is available at https://github.com/marksgraham/ddpm-ood, Comment: Accepted at MICCAI 2023

This paper applies insights from global value chains (GVC)/global production networks (GPN) frameworks to explore the economic geographies brought into being by digital labour platforms. In particular, these perspectives facilitate analyses of power imbalances and value extraction across territories—an under-theorized aspect within platform studies. We theorize this dynamic by introducing the descriptor ‘digital value network’ (DVN): a digitally mediated nexus of platform operations that produce and distribute value between territories, on the basis of labour transactions. Empirically, we draw on a multi-year action research project, assessing the operations of platforms and the experiences of platform workers in 54 countries. Our analysis highlights that platforms as lead firms extend GVC/GPN logics of coordination and drivenness in DVN to (i) optimize production capabilities while externalizing ownership and costs, (ii) accumulate both monetary and non-monetary forms of value, and (iii) concentrate power at the global scale in both existing and new sectors.

The Delta (B.1.617.2) variant was the predominant UK circulating SARS-CoV-2 strain between May and December 2021. How Delta infection compares with previous variants is unknown. This prospective observational cohort study assessed symptomatic adults participating in the app-based COVID Symptom Study who tested positive for SARS-CoV-2 from May 26 to July 1, 2021 (Delta overwhelmingly the predominant circulating UK variant), compared (1:1, age- and sex-matched) with individuals presenting from December 28, 2020 to May 6, 2021 (Alpha (B.1.1.7) the predominant variant). We assessed illness (symptoms, duration, presentation to hospital) during Alpha- and Delta-predominant timeframes; and transmission, reinfection, and vaccine effectiveness during the Delta-predominant period. 3581 individuals (aged 18 to 100 years) from each timeframe were assessed. The seven most frequent symptoms were common to both variants. Within the first 28 days of illness, some symptoms were more common with Delta versus Alpha infection (including fever, sore throat, and headache) and some vice versa (dyspnoea). Symptom burden in the first week was higher with Delta versus Alpha infection; however, the odds of any given symptom lasting ≥ 7 days was either lower or unchanged. Illness duration ≥ 28 days was lower with Delta versus Alpha infection, though unchanged in unvaccinated individuals. Hospitalisation for COVID-19 was unchanged. The Delta variant appeared more (1.49) transmissible than Alpha. Re-infections were low in all UK regions. Vaccination markedly reduced the risk of Delta infection (by 69-84%). We conclude that COVID-19 from Delta or Alpha infections is similar. The Delta variant is more transmissible than Alpha; however, current vaccines showed good efficacy against disease. This research framework can be useful for future comparisons with new emerging variants.

The growth of the platform economy in Belgium, although less rapid than in its neighbouring countries, has triggered heated debates around workers’ legal classification and their protection. The structures for collective employment regulations and strong trade unions under the Ghent system provide some protections for workers, but in the specific context of the platform economy, they do not always translate into fairer working conditions. The lack of guaranteed minimum wage and social security coverage for self-employed workers in the platform economy results in underpaid and unpaid labour. Falling outside the scope of collective sectoral agreements, platform work is also linked to lack of job protection and lack of collective voice. This study by the Fairwork Belgium team sheds light on the working conditions of platform workers and makes practical suggestions for improving them. ispartof: Fairwork Belgium Ratings 2022: Labour Standards in the Platform Economy status: published

Only ten years ago, there were more internet users in countries like France or Germany than in all of Africa put together. But much has changed in a decade. The year 2018 marks the first year in human history in which a majority of the world’s population are now connected to the internet. This mass connectivity means that we have an internet that no longer connects only the world’s wealthy. Workers from Lagos to Johannesburg to Nairobi and everywhere in between can now apply for and carry out jobs coming from clients who themselves can be located anywhere in the world. Digital outsourcing firms can now also set up operations in the most unlikely of places in order to tap into hitherto disconnected labour forces. With CEOs in the Global North proclaiming that ‘location is a thing of the past’ (Upwork, 2018), and governments and civil society in Africa promising to create millions of jobs on the continent, the book asks what this ‘new world of digital work’ means to the lives of African workers. It draws from a year-long fieldwork in South Africa, Kenya, Nigeria, Ghana, and Uganda, with over 200 interviews with participants including gig workers, call and contact centre workers, self-employed freelancers, small-business owners, government officials, labour union officials, and industry experts. Focusing on both platform-based remote work and call and contact centre work, the book examines the job quality implications of digital work for the lives and livelihoods of African workers.

Human anatomy, morphology, and associated diseases can be studied using medical imaging data. However, access to medical imaging data is restricted by governance and privacy concerns, data ownership, and the cost of acquisition, thus limiting our ability to understand the human body. A possible solution to this issue is the creation of a model able to learn and then generate synthetic images of the human body conditioned on specific characteristics of relevance (e.g., age, sex, and disease status). Deep generative models, in the form of neural networks, have been recently used to create synthetic 2D images of natural scenes. Still, the ability to produce high-resolution 3D volumetric imaging data with correct anatomical morphology has been hampered by data scarcity and algorithmic and computational limitations. This work proposes a generative model that can be scaled to produce anatomically correct, high-resolution, and realistic images of the human brain, with the necessary quality to allow further downstream analyses. The ability to generate a potentially unlimited amount of data not only enables large-scale studies of human anatomy and pathology without jeopardizing patient privacy, but also significantly advances research in the field of anomaly detection, modality synthesis, learning under limited data, and fair and ethical AI. Code and trained models are available at: https://github.com/AmigoLab/SynthAnatomy., Comment: 13 pages, 3 figures, 2 tables, accepted at SASHIMI MICCAI 2022

Deep generative models have emerged as promising tools for detecting arbitrary anomalies in data, dispensing with the necessity for manual labelling. Recently, autoregressive transformers have achieved state-of-the-art performance for anomaly detection in medical imaging. Nonetheless, these models still have some intrinsic weaknesses, such as requiring images to be modelled as 1D sequences, the accumulation of errors during the sampling process, and the significant inference times associated with transformers. Denoising diffusion probabilistic models are a class of non-autoregressive generative models recently shown to produce excellent samples in computer vision (surpassing Generative Adversarial Networks), and to achieve log-likelihoods that are competitive with transformers while having fast inference times. Diffusion models can be applied to the latent representations learnt by autoencoders, making them easily scalable and great candidates for application to high dimensional data, such as medical images. Here, we propose a method based on diffusion models to detect and segment anomalies in brain imaging. By training the models on healthy data and then exploring its diffusion and reverse steps across its Markov chain, we can identify anomalous areas in the latent space and hence identify anomalies in the pixel space. Our diffusion models achieve competitive performance compared with autoregressive approaches across a series of experiments with 2D CT and MRI data involving synthetic and real pathological lesions with much reduced inference times, making their usage clinically viable.

In a clinical setting it is essential that deployed image processing systems are robust to the full range of inputs they might encounter and, in particular, do not make confidently wrong predictions. The most popular approach to safe processing is to train networks that can provide a measure of their uncertainty, but these tend to fail for inputs that are far outside the training data distribution. Recently, generative modelling approaches have been proposed as an alternative; these can quantify the likelihood of a data sample explicitly, filtering out any out-of-distribution (OOD) samples before further processing is performed. In this work, we focus on image segmentation and evaluate several approaches to network uncertainty in the far-OOD and near-OOD cases for the task of segmenting haemorrhages in head CTs. We find all of these approaches are unsuitable for safe segmentation as they provide confidently wrong predictions when operating OOD. We propose performing full 3D OOD detection using a VQ-GAN to provide a compressed latent representation of the image and a transformer to estimate the data likelihood. Our approach successfully identifies images in both the far- and near-OOD cases. We find a strong relationship between image likelihood and the quality of a model's segmentation, making this approach viable for filtering images unsuitable for segmentation. To our knowledge, this is the first time transformers have been applied to perform OOD detection on 3D image data. Code is available at github.com/marksgraham/transformer-ood., Comment: Accepted at MIDL 2022 (Oral)

Out-of-distribution detection is crucial to the safe deployment of machine learning systems. Currently, unsupervised out-of-distribution detection is dominated by generative-based approaches that make use of estimates of the likelihood or other measurements from a generative model. Reconstruction-based methods offer an alternative approach, in which a measure of reconstruction error is used to determine if a sample is out-of-distribution. However, reconstruction-based approaches are less favoured, as they require careful tuning of the model's information bottleneck - such as the size of the latent dimension - to produce good results. In this work, we exploit the view of denoising diffusion probabilistic models (DDPM) as denoising autoencoders where the bottleneck is controlled externally, by means of the amount of noise applied. We propose to use DDPMs to reconstruct an input that has been noised to a range of noise levels, and use the resulting multi-dimensional reconstruction error to classify out-of-distribution inputs. We validate our approach both on standard computer-vision datasets and on higher dimension medical datasets. Our approach outperforms not only reconstruction-based methods, but also state-of-the-art generative-based approaches. Code is available at https://github.com/marksgraham/ddpm-ood.

Digital labour platforms have been widely promoted as a solution to the unemployment crisis sparked by the COVID-19 pandemic. However, the pandemic has also highlighted the harms to gig workers-who are exposed either to income loss, or to infection while carrying out essential work, but excluded from labour protections. We examine the COVID-19 policies of 191 platforms in 43 countries to understand how the crisis has shifted the conventions of the gig economy. Using a typology of "fair platform work" we report the introduction of some positive worker protections, but also significant shortfalls, including entrenchment of precarious work as platforms leverage the opportunities arising from the crisis.

ObjectivesDietary supplements may ameliorate SARS-CoV-2 infection, although scientific evidence to support such a role is lacking. We investigated whether users of the COVID-19 Symptom Study app who regularly took dietary supplements were less likely to test positive for SARS-CoV-2 infection.DesignApp-based community survey.Setting445 850 subscribers of an app that was launched to enable self-reported information related to SARS-CoV-2 infection for use in the general population in the UK (n=372 720), the USA (n=45 757) and Sweden (n=27 373).Main exposureSelf-reported regular dietary supplement usage (constant use during previous 3 months) in the first waves of the pandemic up to 31 July 2020.Main outcome measuresSARS-CoV-2 infection confirmed by viral RNA reverse transcriptase PCR test or serology test before 31 July 2020.ResultsIn 372 720 UK participants (175 652 supplement users and 197 068 non-users), those taking probiotics, omega-3 fatty acids, multivitamins or vitamin D had a lower risk of SARS-CoV-2 infection by 14% (95% CI (8% to 19%)), 12% (95% CI (8% to 16%)), 13% (95% CI (10% to 16%)) and 9% (95% CI (6% to 12%)), respectively, after adjusting for potential confounders. No effect was observed for those taking vitamin C, zinc or garlic supplements. On stratification by sex, age and body mass index (BMI), the protective associations in individuals taking probiotics, omega-3 fatty acids, multivitamins and vitamin D were observed in females across all ages and BMI groups, but were not seen in men. The same overall pattern of association was observed in both the US and Swedish cohorts.ConclusionIn women, we observed a modest but significant association between use of probiotics, omega-3 fatty acid, multivitamin or vitamin D supplements and lower risk of testing positive for SARS-CoV-2. We found no clear benefits for men nor any effect of vitamin C, garlic or zinc. Randomised controlled trials are required to confirm these observational findings before any therapeutic recommendations can be made.

Supplemental material, Supplemental_Material for A review of available software for adaptive clinical trial design by Michael John Grayling and Graham Mark Wheeler in Clinical Trials

Many atlases used for brain parcellation are hierarchically organised, progressively dividing the brain into smaller sub-regions. However, state-of-the-art parcellation methods tend to ignore this structure and treat labels as if they are `flat'. We introduce a hierarchically-aware brain parcellation method that works by predicting the decisions at each branch in the label tree. We further show how this method can be used to model uncertainty separately for every branch in this label tree. Our method exceeds the performance of flat uncertainty methods, whilst also providing decomposed uncertainty estimates that enable us to obtain self-consistent parcellations and uncertainty maps at any level of the label hierarchy. We demonstrate a simple way these decision-specific uncertainty maps may be used to provided uncertainty-thresholded tissue maps at any level of the label tree., Comment: To be published in the MICCAI 2020 workshop: Uncertainty for Safe Utilization of Machine Learning in Medical Imaging

The increasing efficiency and compactness of deep learning architectures, together with hardware improvements, have enabled the complex and high-dimensional modelling of medical volumetric data at higher resolutions. Recently, Vector-Quantised Variational Autoencoders (VQ-VAE) have been proposed as an efficient generative unsupervised learning approach that can encode images to a small percentage of their initial size, while preserving their decoded fidelity. Here, we show a VQ-VAE inspired network can efficiently encode a full-resolution 3D brain volume, compressing the data to $0.825\%$ of the original size while maintaining image fidelity, and significantly outperforming the previous state-of-the-art. We then demonstrate that VQ-VAE decoded images preserve the morphological characteristics of the original data through voxel-based morphology and segmentation experiments. Lastly, we show that such models can be pre-trained and then fine-tuned on different datasets without the introduction of bias.

ObjectiveApoC-III (apolipoprotein C-III) glycosylation can predict cardiovascular disease risk. Higher abundance of disialylated (apoC-III2) over monosialylated (apoC-III1) glycoforms is associated with lower plasma triglyceride levels. Yet, it remains unclear whether apoC-III glycosylation impacts TRL (triglyceride-rich lipoprotein) clearance and whether apoC-III antisense therapy (volanesorsen) affects distribution of apoC-III glycoforms. Approach and Results: To measure the abundance of human apoC-III glycoforms in plasma over time, human TRLs were injected into wild-type mice and mice lacking hepatic TRL clearance receptors, namely HSPGs (heparan sulfate proteoglycans) or both LDLR (low-density lipoprotein receptor) and LRP1 (LDLR-related protein 1). ApoC-III was more rapidly cleared in the absence of HSPG (t1/2=25.4 minutes) than in wild-type animals (t1/2=55.1 minutes). In contrast, deficiency of LDLR and LRP1 (t1/2=56.1 minutes) did not affect clearance of apoC-III. After injection, a significant increase in the relative abundance of apoC-III2 was observed in HSPG-deficient mice, whereas the opposite was observed in mice lacking LDLR and LRP1. In patients, abundance of plasma apoC-III glycoforms was assessed after placebo or volanesorsen administration. Volanesorsen treatment correlated with a statistically significant 1.4-fold increase in the relative abundance of apoC-III2 and a 15% decrease in that of apoC-III1. The decrease in relative apoC-III1 abundance was strongly correlated with decreased plasma triglyceride levels in patients.ConclusionsOur results indicate that HSPGs preferentially clear apoC-III2. In contrast, apoC-III1 is more effectively cleared by LDLR/LRP1. Clinically, the increase in the apoC-III2/apoC-III1 ratio on antisense lowering of apoC-III might reflect faster clearance of apoC-III1 because this metabolic shift associates with improved triglyceride levels.

Figures S1. a and b show the correlation matrices for the support and barrier items. Darker colors indicate greater correlation. Bolded items are significantly correlated with each other at (minimally) a significance value of less than 0.05, with most being lower than .01. Table S1. Mapping of implementation items to the taxonomy in Couch et al. (2015). (DOCX 408 kb)

We propose the construction of a Digital Knowledge Economy Index, quantified by way of measuring content creation and participation through digital platforms, namely the code sharing platform GitHub, the crowdsourced encyclopaedia Wikipedia, and Internet domain registrations and estimating a fifth sub-index for the World Bank Knowledge Economy Index for year 2012. This approach complements conventional data sources such as national statistics and expert surveys and helps reflect the underlying digital content creation, capacities, and skills of the population. An index that combines traditional and novel data sources can provide a more revealing view of the status of the world’s digital knowledge economy and highlight where the (un)availability of digital resources may actually reinforce inequalities in the age of data.

It has been observed that low redshift early-type galaxies can be separated into slow and fast rotators according to a proxy of specific stellar angular momentum, $\lambda_{R_e}$. Detailed studies of a handful of nearby clusters have shown that slow rotators are generally found at the centres of clusters where the number density is highest, whereas the fast rotators trace the trend followed by early-type galaxies of increasing in number with local density. In this paper, we study the environmental distribution of slow and fast rotators using the stellar kinematics of about 3900 galaxies from the Sloan Digital Sky Survey's Mapping Nearby Galaxies at Apache Point Observatory survey. For galaxies in groups closer than $z=0.08$ that are not observed with MaNGA but satisfy the necessary conditions for slow rotators, we visually assign slow/fast rotator classifications to obtain a complete sample. Our final catalogue contains about 900 groups of five or more members. We observe the kinematic morphology-density (kT-$\Sigma$) relation for each group and find an increasing fraction of massive slow rotators with increasing number density. We provide evidence suggesting that the observed lack of trends in angular momentum with environment at fixed stellar mass is partly because the maximum density varies between clusters, and that the locations of massive slow rotators are strongly correlated with peak densities in galaxy groups and clusters. We conclude that the (projected) number density relative to the cluster peak is more fundamental than the absolute number density in influencing the abundance of slow rotators. We find that the kT-$\Sigma$ relation does exist at fixed stellar mass, and we rule out the hypothesis that the kT-$\Sigma$ relation is a result of dynamical friction alone, instead arguing that massive slow rotators grow hierarchically in tandem with their host clusters., Comment: Submitted to Astronomy & Astrophysics on October 6th 2019. This paper is the third in a series of three papers

Fibrosis is the major determinant of morbidity and mortality in patients with nonalcoholic steatohepatitis (NASH)but has no approved pharmacotherapy in part because of incomplete understanding of its pathogenic mechanisms.Here, we report that hepatocyte Notch activity tracks with disease severity and treatment response in patients withNASH and is similarly increased in a mouse model of diet-induced NASH and liver fibrosis. Hepatocyte-specificNotch loss-of-function mouse models showed attenuated NASH-associated liver fibrosis, demonstrating causalityto obesity-induced liver pathology. Conversely, forced activation of hepatocyte Notch induced fibrosis in bothchow- and NASH diet–fed mice by increasing Sox9-dependent Osteopontin (Opn) expression and secretion from hepatocytes, which activate resident hepatic stellate cells. In a cross-sectional study, we found that OPN explainsthe positive correlation between liver Notch activity and fibrosis stage in patients. Further, we developed a Notchinhibitor [Nicastrin antisense oligonucleotide (Ncst ASO)] that reduced fibrosis in NASH diet–fed mice. In summary,these studies demonstrate the pathological role and therapeutic accessibility of the maladaptive hepatocyteNotch response in NASH-associated liver fibrosis.

This paper introduces the Fairwork Foundation, a research initiative that is also developing an intervention around the quality of work on digital labour platforms. Lacking the ability to collectively bargain, many of these workers have little ability to negotiate wages or working conditions with their employers who are often on the other side of the world. As a result of this new global market for work, many workers have jobs characterized by long and irregular hours, low income, and high stress. Across India and South Africa, there are challenges for workers across a range of issues, including: pay, conditions, contracts, management, and representation. The results of the fieldwork are being used to rank and compare platforms as part of the ongoing ‘work in progress’ of the Fairwork project, a research initiative that is developing an intervention to improve the quality of work on digital platforms., Weizenbaum Conference

ISTAR is an integrated project support environment produced by Imperial Software Technology, Ltd. This evaluation of ISTAR is intended for software technologists considering the adoption of an integrated project support environment. Researchers and others interested in environments and evaluation methods will also benefit from this report.

Engaging undergraduate students in scientific research promises substantial benefits, but it is not accessible to all students and is rarely implemented early in college education, when it will have the greatest impact. An inclusive Research Education Community (iREC) provides a centralized scientific and administrative infrastructure enabling engagement of large numbers of students at different types of institutions. The Science Education Alliance-Phage Hunters Advancing Genomics and Evolutionary Science (SEA-PHAGES) is an iREC that promotes engagement and continued involvement in science among beginning undergraduate students. The SEA-PHAGES students show strong gains correlated with persistence relative to those in traditional laboratory courses regardless of academic, ethnic, gender, and socioeconomic profiles. This persistent involvement in science is reflected in key measures, including project ownership, scientific community values, science identity, and scientific networking.

Human geographies are infused with information. Information is created, processed, and used in places. It is stored in places, moves across places, and ultimately annotates and augments places.

In this paper a Minkowski analogue of the Euclidean medial axis of a closed and smooth plane curve is introduced. Its generic local configurations are studied and the types of shocks that occur on these are also determined.

Cette recherche part du constat d’une faible participation à, et d’un maigre contenu dans Wikipedia Arabe et questionne les obstacles à un contenu plus riche. La problématique de la recherche est double: d’abord proposer une cartographie numérique de Wikipedia dans la région du Moyen-Orient et Afrique du Nord. Cette cartographie informe d’une part sur la représentation des participants à Wikipedia Arabe, et d’autre part, sur la contribution à la distribution du savoir arabe et de son étendue. Deuxièmement, la recherche s’intéresse au contenu de Wikipedia arabe et de son mode de production. Plus particulièrement, les questions de recherche portent sur la participation à la production de contenu de Wikipedia Arabe, les barrières à l’entrée dans Wikipedia Arabe, les dynamiques de négociation et de consensus entre éditeurs et administrateurs Wikipediens de la région et les solutions envisagées pour améliorer et encourager la production de contenu. La démarche méthodologique est principalement qualitative.So far Arabic content on Wikipedia Arabic is very weak; this paper questions the barriers that prevent contributors to Wikipedia Arabic from being more active in content production on Wikipedia Arabic. The research problematic is twofold: first, it suggests a digital cartography of Wikipedia in the MENA region that informs about those who represent Wikipedia Arabic and thus, contribute to the Arab knowledge distribution. Second, while adopting a qualitative approach, the research looks at the content of Arabic Wikipedia. It questions its production methods, contribution barriers, dynamics of negotiation between Wikipedians, and strategies of consensus between editors and admins, as well as the envisioned solutions to improve and promote Wikipedia Arabic content production.

This chapter outlines how one might utilize the massive amounts of web-based, geographically-referenced digital social data for geographical research. Because much of these data are user-generated and produced through social media platforms, we also focus on the pitfalls associated with such sources and the benefits of a mixed methods approach to these data. Not only can digital social data be mapped for visual analysis, it is also useful to use a range of quantitative methods to understand relationships between different subsets of the data. In addition, closer, systematic readings via qualitative methods of social data provides insights of particular people’s perceptions and experiences of the world around them. Thus, while making maps is often the starting point for geographers working with this kind of research, it is rarely the end point.

Objective To review the long-term follow-up data from the authors’ institutional experience of 62 patients with locally advanced breast cancer (LABC) treated with a uniform multimodality regimen. The authors determined the rate of breast preservation, the disease-free and overall survival, and the factors associated with locoregional and distant recurrent disease. Summary Background Data It remains a challenge to achieve local and distant control of LABC. Over the last decade, preoperative or neoadjuvant chemotherapy has emerged as the standard of care for these patients. Successful tumor downstaging has been associated with increased rates of breast-conserving therapy (BCT), but the overall effect on long-term survival remains to be seen. Methods This study examines a cohort of 62 patients with LABC treated at the authors’ institution from 1992 to 1998. The uniform treatment regimen consisted of neoadjuvant doxorubicin (Adriamycin), followed by operation (BCT if sufficient clinical downstaging), followed by non-cross-resistant cyclophosphamide/methotrexate/5-fluorouracil, followed by radiation therapy. Treatment was both dose-intensive and time-intensive, with a total treatment time of 32 to 35 weeks. Results In this patient population, the median age was 44 years, with approximately two thirds white patients and one third African-American. Eighty-two percent of patients were clinical stage III at presentation, 13 patients had T4d inflammatory cancers, and 3 patients were stage IV at diagnosis. Eighty-four percent of patients demonstrated a significant clinical response to doxorubicin. Twenty-eight patients had sufficient clinical downstaging to attempt BCT, and 22 (45%) of 49 noninflammatory patients underwent successful BCT. Pathologic complete response was seen in 15% of patients. Median follow-up for the cohort was 70 months. The local recurrence rate was 14%, including two ipsilateral breast tumor recurrences (10%) in the BCT patients. Seven (12%) patients developed a new primary cancer in the contralateral breast. Distant metastases occurred in 18 (31%) patients, and the 5-year overall survival rate for the cohort was 76%. Furthermore, in the patients who underwent an attempt at BCT, the survival rate was 96% at 5 years. Conclusions Dose-intensive and time-intensive multimodality neoadjuvant therapy was successfully administered to a mixed racial group over shortened times. Patients who had sufficient clinical downstaging to allow BCT have the best long-term outcome. Patients who required mastectomy are at a higher risk of relapse, as well as the development of new contralateral cancers, yet have 5-year survival rates of over 50%.

Technological progress alone is insufficient in the E&P industry today given the wider administrative, legal and organisational challenges that E&P development Projects currently have to contend with. In particular, risk management has become fundamental to both operator and service provider companies. This paper presents two E&P case studies that illustrate how the use of collaboration and data visualisation tools provided significant assistance and support to E&P risk management activities. The first case study concerns a deep offshore E&P development project where the risk management activity had become virtually impossible to coordinate because of a very aggressive project execution environment. The implementation of collaboration technology for the project's executive management team established and sustained risk visibility at the appropriate level, whilst allowing all risk-related information to be seamlessly integrated with the rest of the project execution plan. The second case study concerns the collection and monitoring of safety and risk performance indicators (RPIs) at operational E&P sites of a major oil company. We will describe how analysis and feasibility work led to the design and development of a methodological tool capable of collecting and organising a diversified set of RPIs and then presenting them in an innovative graphical form for analysis and prioritisation. The added value for E&P projects is evident: risk management without risk.

An overview of an object-oriented class library, CFEL, for finite element modelling is presented. The structure and class hierarchies of CFEL are presented and how the library forms part of a larger computational modelling class framework, CML. The CML framework is a move towards a general framework for geometric computational modelling which unifies traditionally separate disciplines such as stress analysis and fluid flow. CML consists of the following set of core libraries for modelling: i) data structures, ii) element and structure geometric modelling, iii) material and geometric characteristics and iv) solid and fluid finite element modelling, of which CFEL is a member.

The SQL ADA Module Description Language (SAMeDL) is a language for describing information services to be provided to ADA application programs by SQL database management systems. This report shows how the SAMeDL can be adapted and extended to provide services to applications needing advanced features (e.g., dynamic SQL, or using non-ANSI standard data types (decimal, date) or having other unusual requirements. It also contains short descriptions of some implementation details.

The Software Engineering Institute (SEI) hosted the Requirements Engineering and Analysis Workshop in Pittsburgh, Pennsylvania, on March 12-14, 1991. The intention of the workshop was to focus discussion on issues and activities that could help the Department of Defense (DoD) to deal more effectively with the requirements of mission-critical systems. The SEI workshop built upon work performed previously at the Requirements Engineering and Rapid Prototyping Workshop held by the U.S. Army Communications-Electronics Command (CECOM) Center for Software Engineering in Eatontown, New Jersey, on November 14-16, 1989. The workshop participants were divided into four working groups: Requirements Engineering Process and Products, Requirements Volatility, Requirements Elicitation, and Requirements Engineering Techniques and Tools. A summary of the findings of each working group follows.

The SEI was asked by the ADA Joint Program Office to investigate the problem of interfacing programs written in ADA with database management systems implementing the SQL database language. The authors decided to concentrate on a description of the problems involved in producing an interface that would be worthy of becoming a standard. This document is meant to assist the reader in answering the question "What constitutes a good interface between ADA and SQL?" The document should be useful both in the production of a standard and in the analysis of any proposed standard.

The SEI was asked by the ADA Joint Program Office to investigate the problem of interfacing programs written in ADA with database management systems implementing the SQL database language. The authors decided to concentrate on a description of the problems involved in producing an interface that would be worthy of becoming a standard. This document is meant to assist the reader in answering the question "What constitutes a good interface between ADA and SQL?" The document should be useful both in the production of a standard and in the analysis of any proposed standard.