No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
An HPC-Driven Data Science Platform to Speed-up Time Series Data Analysis of Patients with the Acute Respiratory Distress Syndrome
Abstract
An increasing number of data science approaches that take advantage of deep learning in computational medicine and biomedical engineering require parallel and scalable algorithms using High-Performance Computing systems. Especially computational methods for analysing clinical datasets that consist of multivariate time series data can benefit from High-Performance Computing when applying computing-intensive Recurrent Neural Networks. This paper proposes a dynamic data science platform consisting of modular High-Performance Computing systems using accelerators for innovative Deep Learning algorithms to speed-up medical applications that take advantage of large biomedical scientific databases. This platform’s core idea is to train a set of Deep Learning models very fast to easily combine and compare the different Deep Learning models’ forecast (out-of-sample) performance to their past (in-sample) performance. Considering that this enables a better understanding of what Deep Learning models can be useful to apply to specific medical datasets, our case study leverages the three data science methods Gated Recurrent Units, one-dimensional convolutional layers, and their combination. We validate our approach using the open MIMIC-III database in a case study that assists in understanding, diagnosing, and treating a specific condition that affects Intensive Care Unit patients, namely Acute Respiratory Distress Syndrome.
... Their work highlights the speed-up that can be achieved by making use of HPC, especially in situations where many trials need to be performed with minute changes in order to find the optimal parameter combination that produces the best results. This paper describes the process by which an ML and data science platform that takes advantage of Modular Supercomputing Architecture (MSA) available from the Jülich Supercomputing Centre (JSC) is used to build a surrogate model of the NPS with the intention of implementing it for streamlined ARDS-diagnosis support [28][29][30]. In order to achieve this primary goal, several steps need to be completed as follows: [32]. ...
... These surrogates benefit greatly from the high accuracy of the mechanistic models they emulate, while avoiding the computation overhead associated with equilibrating multiple complex differential equations. This aspect coupled with the use of a pre-established HPC-enabled data science and ML platform that was validated in previously published work represent the core innovations of the research described in this manuscript [28,29]. In this way, the HPC resources are instrumental to the accelerated development and testing of the surrogate. ...
... Their work highlights the speed-up that can be achieved by making use of HPC, especially in situations where many trials need to be performed with minute changes in order to find the optimal parameter combination that produces the best results. This paper describes the process by which an ML and data science platform that takes advantage of Modular Supercomputing Architecture (MSA) available from the Jülich Supercomputing Centre (JSC) is used to build a surrogate model of the NPS with the intention of implementing it for streamlined ARDS-diagnosis support [28][29][30]. In order to achieve this primary goal, several steps need to be completed as follows: [32]. ...
... These surrogates benefit greatly from the high accuracy of the mechanistic models they emulate, while avoiding the computation overhead associated with equilibrating multiple complex differential equations. This aspect coupled with the use of a pre-established HPC-enabled data science and ML platform that was validated in previously published work represent the core innovations of the research described in this manuscript [28,29]. In this way, the HPC resources are instrumental to the accelerated development and testing of the surrogate. ...
Acute Respiratory Distress Syndrome (ARDS) is a condition that endangers the lives of many Intensive Care Unit patients through gradual reduction of lung function. Due to its heterogeneity, this condition has been difficult to diagnose and treat, although it has been the subject of continuous research, leading to the development of several tools for modeling disease progression on the one hand, and guidelines for diagnosis on the other, mainly the “Berlin Definition”. This paper describes the development of a deep learning-based surrogate model of one such tool for modeling ARDS onset in a virtual patient: the Nottingham Physiology Simulator. The model-development process takes advantage of current machine learning and data-analysis techniques, as well as efficient hyperparameter-tuning methods, within a high-performance computing-enabled data science platform. The lightweight models developed through this process present comparable accuracy to the original simulator (per-parameter R2 > 0.90). The experimental process described herein serves as a proof of concept for the rapid development and dissemination of specialised diagnosis support systems based on pre-existing generalised mechanistic models, making use of supercomputing infrastructure for the development and testing processes and supported by open-source software for streamlined implementation in clinical routines.
... The nodes can vary in number, and they incorporate memory units alongside comprehensive operating systems dependent on what specifications best suit their intended use [20]. The main system components include individual machines interlinked via fast interconnects, in addition to software features geared towards enabling maximized performance throughput during parallel execution operations, ensuring prompt task delivery [21]. ...
This study focuses on the challenge of developing abstract models to differentiate various cloud resources. It explores the advancements in cloud products that offer specialized services to meet specific external needs. The study proposes a new approach to request processing in clusters, improving downtime, load distribution, and overall performance. A comparison of three clustering approaches is conducted: local single cluster, local multiple clusters, and multiple cloud clusters. Performance, scalability, fault tolerance, resource allocation, availability , and cost-effectiveness are evaluated through experiments with 50 requests. All three approaches achieve a 100% success rate, but processing times vary. The local single cluster has the longest duration, while the local multiple clusters and multiple cloud clusters perform better and offer faster processing, scalability, fault tolerance, and availability. From a cost perspective, the local single cluster and local multiple clusters incur capital and operational expenses, while the multiple cloud clusters follow a pay-as-you-go model. Overall, the local multiple clusters and multiple cloud clusters outperform the local single cluster in terms of performance , scalability, fault tolerance, resource allocation, availability, and cost-effectiveness. These findings provide valuable insights for selecting appropriate clustering strategies in cloud environments.
Many artificially intelligent systems solve complex health- and agriculture-related problems that require great computational power. Such systems are used for tracking medical records, genome sequence analysis, image-based plant disease detection, food supply chain traceability, and photosynthesis simulation. Massively parallel computers (MPCs) are among those used to solve these computation-intensive problems. MPCs comprise a million nodes; connecting such a large number of nodes is a daunting task. Therefore, hierarchical interconnection networks (HINs) have been introduced to solve this problem. A midimew-connected torus network (MTN) is a HIN that has basic modules (BM) as torus networks that are connected hierarchically by midimew links. This paper presents the performance of MTNs in terms of static topological parameters and cost-effectiveness, as measured through simulations. An MTN was compared with other networks, including mesh, torus, TESH, TTN, MMN, and TFBN. The results showed that our MTN had a low diameter with a high bisection width and arc connectivity. In addition, our MTN had a high cost–performance trade-off factor (CPTF), a high cost-effective factor (CEF), low packing density, and moderate message-traffic density with marginally higher costs, as compared to other networks, due to wire complexity. However, our MTN provided better bandwidth with higher static fault tolerance. Therefore, MTNs are suggested for further evaluation of the effective implementation of MPCs.
The COVID-19 pandemic shed light on the need for quick diagnosis tools in healthcare, leading to the development of several algorithmic models for disease detection. Though these models are relatively easy to build, their training requires a lot of data, storage, and resources, which may not be available for use by medical institutions or could be beyond the skillset of the people who most need these tools. This paper describes a data analysis and machine learning platform that takes advantage of high-performance computing infrastructure for medical diagnosis support applications. This platform is validated by re-training a previously published deep learning model (COVID-Net) on new data, where it is shown that the performance of the model is improved through large-scale hyperparameter optimisation that uncovered optimal training parameter combinations. The per-class accuracy of the model, especially for COVID-19 and pneumonia, is higher when using the tuned hyperparameters (healthy: 96.5%; pneumonia: 61.5%; COVID-19: 78.9%) as opposed to parameters chosen through traditional methods (healthy: 93.6%; pneumonia: 46.1%; COVID-19: 76.3%). Furthermore, training speed-up analysis shows a major decrease in training time as resources increase, from 207 min using 1 node to 54 min when distributed over 32 nodes, but highlights the presence of a cut-off point where the communication overhead begins to affect performance. The developed platform is intended to provide the medical field with a technical environment for developing novel portable artificial-intelligence-based tools for diagnosis support.
Acute Respiratory Distress Syndrome (ARDS), also known as noncardiogenic pulmonary edema, is a severe condition that affects around one in ten-thousand people every year with life-threatening consequences. Its pathophysiology is characterized by bronchoalveolar injury and alveolar collapse (i.e., atelectasis), whereby its patient diagnosis is based on the so-called ‘Berlin Definition‘. One common practice in Intensive Care Units (ICUs) is to use lung recruitment manoeuvres (RMs) in ARDS to open up unstable, collapsed alveoli using a temporary increase in transpulmonary pressure. Many RMs have been proposed, but there is also confusion regarding the optimal way to achieve and maintain alveolar recruitment in ARDS. Therefore, the best solution to prevent lung damages by ARDS is to identify the onset of ARDS which is still a matter of research. Determining ARDS disease onset, progression, diagnosis, and treatment required algorithmic support which in turn raises the demand for cutting-edge computing power. This paper thus describes several different data science approaches to better understand ARDS, such as using time series analysis and image recognition with deep learning methods and mechanistic modelling using a lung simulator. In addition, we outline how High-Performance Computing (HPC) helps in both cases. That also includes porting the mechanistic models from serial MatLab approaches and its modular supercomputer designs. Finally, without losing sight of discussing the datasets, their features, and their relevance, we also include broader selected lessons learned in the context of ARDS out of our Smart Medical Information Technology for Healthcare (SMITH) research project. The SMITH consortium brings together technologists and medical doctors of nine hospitals, whereby the ARDS research is performed by our Algorithmic Surveillance of ICU (ASIC) patients team. The paper thus also describes how it is essential that HPC experts team up with medical doctors that usually lack the technical and data science experience and contribute to the fact that a wealth of data exists, but ARDS analysis is still slowly progressing. We complement the ARDS findings with selected insights from our Covid-19 research under the umbrella of the European Open Science Cloud (EOSC) fast track grant, a very similar application field.
The novel coronavirus 2019 (COVID-19) is a respiratory syndrome that resembles pneumonia. The current diagnostic procedure of COVID-19 follows reverse-transcriptase polymerase chain reaction (RT-PCR) based approach which however is less sensitive to identify the virus at the initial stage. Hence, a more robust and alternate diagnosis technique is desirable. Recently, with the release of publicly available datasets of corona positive patients comprising of computed tomography (CT) and chest X-ray (CXR) imaging; scientists, researchers and healthcare experts are contributing for faster and automated diagnosis of COVID-19 by identifying pulmonary infections using deep learning approaches to achieve better cure and treatment. These datasets have limited samples concerned with the positive COVID-19 cases, which raise the challenge for unbiased learning. Following from this context, this article presents the random oversampling and weighted class loss function approach for unbiased fine-tuned learning (transfer learning) in various state-of-the-art deep learning approaches such as baseline ResNet, Inception-v3, Inception ResNet-v2, DenseNet169, and NASNetLarge to perform binary classification (as normal and COVID-19 cases) and also multi-class classification (as COVID-19, pneumonia, and normal case) of posteroanterior CXR images. Accuracy, precision, recall, loss, and area under the curve (AUC) are utilized to evaluate the performance of the models. Considering the experimental results, the performance of each model is scenario dependent; however, NASNetLarge displayed better scores in contrast to other architectures, which is further compared with other recently proposed approaches. This article also added the visual explanation to illustrate the basis of model classification and perception of COVID-19 in CXR images.
Purpose
Acute respiratory distress syndrome (ARDS) is a serious respiratory condition with high mortality and associated morbidity. The objective of this study is to develop and evaluate a novel application of gradient boosted tree models trained on patient health record data for the early prediction of ARDS.
Materials and methods
9919 patient encounters were retrospectively analyzed from the Medical Information Mart for Intensive Care III (MIMIC-III) data base. XGBoost gradient boosted tree models for early ARDS prediction were created using routinely collected clinical variables and numerical representations of radiology reports as inputs. XGBoost models were iteratively trained and validated using 10-fold cross validation.
Results
On a hold-out test set, algorithm classifiers attained area under the receiver operating characteristic curve (AUROC) values of 0.905 when tested for the detection of ARDS at onset and 0.827, 0.810, and 0.790 for the prediction of ARDS at 12-, 24-, and 48-h windows prior to onset, respectively.
Conclusion
Supervised machine learning predictions may help predict patients with ARDS up to 48 h prior to onset.
With the wide application of computer technology, medical health data has also increased dramatically, and data-driven medical big data analysis methods have emerged as the times require, providing assistance for intelligent identification of medical health. However, due to the mixed medical big data format, many incomplete records, and a lot of noise, it is still difficult to analyze medical big data. Traditional machine learning methods can’t effectively mine the rich information contained in medical big data, while deep learning builds a hierarchical model by simulating the human brain. It has powerful automatic feature extraction, complex model construction and efficient feature expression, and more important. It is a deep learning method that extracts features from the bottom to the top level from the original medical image data. Therefore, this paper constructs a data analysis model based on deep learning for medical images and transcripts, and is used for intelligent identification and diagnosis of diseases. The model uses massive medical big data to select and optimize model parameters, and automatically learns the pathological analysis process of doctors or medical researchers through the model, and finally intelligently conducts disease judgment and effective decision based on the analysis results of medical big data. The experimental results show that the method can analyze the medical big data, and can realize the early diagnosis of the disease. At the same time, it can analyze the physical health status according to the patient’s physical examination records and predict the risk of a certain disease in the future. Greatly reduce the work pressure of doctors or medical researchers and improve their work efficiency.
Introduction:
This article is part of the Focus Theme of Methods of Information in Medicine on the German Medical Informatics Initiative. "Smart Medical Information Technology for Healthcare (SMITH)" is one of four consortia funded by the German Medical Informatics Initiative (MI-I) to create an alliance of universities, university hospitals, research institutions and IT companies. SMITH's goals are to establish Data Integration Centers (DICs) at each SMITH partner hospital and to implement use cases which demonstrate the usefulness of the approach.
Objectives:
To give insight into architectural design issues underlying SMITH data integration and to introduce the use cases to be implemented.
Governance and policies:
SMITH implements a federated approach as well for its governance structure as for its information system architecture. SMITH has designed a generic concept for its data integration centers. They share identical services and functionalities to take best advantage of the interoperability architectures and of the data use and access process planned. The DICs provide access to the local hospitals' Electronic Medical Records (EMR). This is based on data trustee and privacy management services. DIC staff will curate and amend EMR data in the Health Data Storage.
Methodology and architectural framework:
To share medical and research data, SMITH's information system is based on communication and storage standards. We use the Reference Model of the Open Archival Information System and will consistently implement profiles of Integrating the Health Care Enterprise (IHE) and Health Level Seven (HL7) standards. Standard terminologies will be applied. The SMITH Market Place will be used for devising agreements on data access and distribution. 3LGM2 for enterprise architecture modeling supports a consistent development process.The DIC reference architecture determines the services, applications and the standardsbased communication links needed for efficiently supporting the ingesting, data nourishing, trustee, privacy management and data transfer tasks of the SMITH DICs. The reference architecture is adopted at the local sites. Data sharing services and the market place enable interoperability.
Use cases:
The methodological use case "Phenotype Pipeline" (PheP) constructs algorithms for annotations and analyses of patient-related phenotypes according to classification rules or statistical models based on structured data. Unstructured textual data will be subject to natural language processing to permit integration into the phenotyping algorithms. The clinical use case "Algorithmic Surveillance of ICU Patients" (ASIC) focusses on patients in Intensive Care Units (ICU) with the acute respiratory distress syndrome (ARDS). A model-based decision-support system will give advice for mechanical ventilation. The clinical use case HELP develops a "hospital-wide electronic medical record-based computerized decision support system to improve outcomes of patients with blood-stream infections" (HELP). ASIC and HELP use the PheP. The clinical benefit of the use cases ASIC and HELP will be demonstrated in a change of care clinical trial based on a step wedge design.
Discussion:
SMITH's strength is the modular, reusable IT architecture based on interoperability standards, the integration of the hospitals' information management departments and the public-private partnership. The project aims at sustainability beyond the first 4-year funding period.
Patients in hospitals, particularly in critical care, are susceptible to many complications affecting morbidity and mortality. Digitized clinical data in electronic medical records can be effectively used to develop machine learning models to identify patients at risk of complications early and provide prioritized care to prevent complications. However, clinical data from heterogeneous sources within hospitals pose significant modeling challenges. In particular, unstructured clinical notes are a valuable source of information containing regular assessments of the patient's condition but contain inconsistent abbreviations and lack the structure of formal documents. Our contributions in this paper are twofold. First, we present a new preprocessing technique for extracting features from informal clinical notes that can be used in a classification model to identify patients at risk of developing complications. Second, we explore the use of collective matrix factorization, a multi-view learning technique, to model heterogeneous clinical data - text-based features in combination with other measurements, such as clinical investigations, comorbidites, and demographic data. We present a detailed case study on postoperative respiratory failure using more than 700 patient records from the MIMIC II database. Our experiments demonstrate the efficacy of our preprocessing technique in extracting discriminatory features from clinical notes as well as the benefits of multi-view learning to combine clinical measurements with text data for predicting complications.
PurposeTo improve the outcome of the acute respiratory distress syndrome (ARDS), one needs to identify potentially modifiable factors associated with mortality. Methods
The large observational study to understand the global impact of severe acute respiratory failure (LUNG SAFE) was an international, multicenter, prospective cohort study of patients with severe respiratory failure, conducted in the winter of 2014 in a convenience sample of 459 ICUs from 50 countries across five continents. A pre-specified secondary aim was to examine the factors associated with outcome. Analyses were restricted to patients (93.1 %) fulfilling ARDS criteria on day 1–2 who received invasive mechanical ventilation. Results2377 patients were included in the analysis. Potentially modifiable factors associated with increased hospital mortality in multivariable analyses include lower PEEP, higher peak inspiratory, plateau, and driving pressures, and increased respiratory rate. The impact of tidal volume on outcome was unclear. Having fewer ICU beds was also associated with higher hospital mortality. Non-modifiable factors associated with worsened outcome from ARDS included older age, active neoplasm, hematologic neoplasm, and chronic liver failure. Severity of illness indices including lower pH, lower PaO2/FiO2 ratio, and higher non-pulmonary SOFA score were associated with poorer outcome. Of the 578 (24.3 %) patients with a limitation of life-sustaining therapies or measures decision, 498 (86.0 %) died in hospital. Factors associated with increased likelihood of limitation of life-sustaining therapies or measures decision included older age, immunosuppression, neoplasia, lower pH and increased non-pulmonary SOFA scores. Conclusions
Higher PEEP, lower peak, plateau, and driving pressures, and lower respiratory rate are associated with improved survival from ARDS.Trial Registration: ClinicalTrials.gov NCT02010073.
Many multivariate time series data in practical applications, such as health care, geoscience, and biology, are characterized by a variety of missing values. It has been noted that the missing patterns and values are often correlated with the target labels, a.k.a., missingness is informative, and there is significant interest to explore methods which model them for time series prediction and other related tasks. In this paper, we develop novel deep learning models based on Gated Recurrent Units (GRU), a state-of-the-art recurrent neural network, to handle missing observations. Our model takes two representations of missing patterns, i.e., masking and time duration, and effectively incorporates them into a deep model architecture so that it not only captures the long-term temporal dependencies in time series, but also utilizes the missing patterns to improve the prediction results. Experiments of time series classification tasks on real-world clinical datasets (MIMIC-III, PhysioNet) and synthetic datasets demonstrate that our models achieve state-of-art performance on these tasks and provide useful insights for time series with missing values.
MIMIC-III (‘Medical Information Mart for Intensive Care’) is a large, single-center database comprising information relating to patients admitted to critical care units at a large tertiary care hospital. Data includes vital signs, medications, laboratory measurements, observations and notes charted by care providers, fluid balance, procedure codes, diagnostic codes, imaging reports, hospital length of stay, survival data, and more. The database supports applications including academic and industrial research, quality improvement initiatives, and higher education coursework.
Computer simulation models could play a key role in developing novel therapeutic strategies for patients with chronic obstructive pulmonary disease (COPD) if they can be shown to accurately represent the pathophysiological characteristics of individual patients.
We evaluated the capability of a computational simulator to reproduce the heterogeneous effects of COPD on alveolar mechanics as captured in a number of different patient datasets.
Our results show that accurately representing the pathophysiology of individual COPD patients necessitates the use of simulation models with large numbers (up to 200) of compartments for gas exchange. The tuning of such complex simulation models 'by hand' to match patient data is not feasible, and thus we present an automated approach based on the use of global optimization algorithms and high-performance computing. Using this approach, we are able to achieve extremely close matches between the simulator and a range of patient data including PaO2, PaCO2, pulmonary deadspace fraction, pulmonary shunt fraction, and ventilation/perfusion ( /Q) curves. Using the simulator, we computed combinations of ventilator settings that optimally manage the trade-off between ensuring adequate gas exchange and minimizing the risk of ventilator-associated lung injury for an individual COPD patient.
Our results significantly strengthen the credibility of computer simulation models as research tools for the development of novel management protocols in COPD and other pulmonary disease states.
Direct comparison of the relative efficacy of different recruitment maneuvers
(RMs) for patients with acute respiratory distress syndrome (ARDS) via clinical trials is
difficult, due to the heterogeneity of patient populations and disease states, as well as a variety
of practical issues. There is also significant uncertainty regarding the minimum values of
positive end expiratory pressure (PEEP) required to ensure maintenance of effective lung
recruitment using RMs. We used patient-specific computational simulation to analyze how
three different RMs act to improve physiological responses, and investigate how different
levels of PEEP contribute to maintaining effective lung recruitment.
The Berlin definition divides acute respiratory distress syndrome (ARDS) into 3 severity categories. The relationship between these categories and pulmonary microvascular permeability as well as extravascular lung water content, which is the hallmark of lung pathophysiology, remains to be elucidated. The aim of this study was to evaluate the relationship between extravascular lung water, pulmonary vascular permeability, and the severity categories as defined by the Berlin definition, and to confirm the associated predictive validity for severity.
The extravascular lung water index (EVLWi) and pulmonary vascular permeability index (PVPI) were measured using a transpulmonary thermodilution method for 3 consecutive days in 195 patients with an EVLWi of >10 mL/kg and who fulfilled the Berlin definition of ARDS. Collectively, these patients were seen at 23 ICUs. Using the Berlin definition, patients were classified into 3 categories: mild, moderate, severe.
Compared to patients with mild ARDS, patients with moderate and severe ARDS had higher Acute Physiology and Chronic Health Evaluation II and sequential organ failure assessment scores on the day of enrollment. Patients with severe ARDS had higher EVLWi (mild, 16.1; moderate, 17.2; severe, 19.1; P < 0.05) and PVPI (2.7; 3.0; 3.2; P < 0.05). When categories were defined by the minimum PaO2/FIO2 ratio observed during the study period, the 28-day mortality rate increased with severity categories: moderate, odds ratio: 3.125 relative to mild; and severe, odds: 4.167 relative to mild. On independent evaluation of 495 measurements from 195 patients over 3 days, negative and moderate correlations were observed between EVLWi and the PaO2/FIO2 ratio (r = -0.355, P< 0.001) as well as between PVPI and the PaO2/FIO2 ratio (r = -0.345, P < 0.001). ARDS severity was associated with an increase in EVLWi with the categories (mild, 14.7; moderate, 16.2; severe, 20.0; P < 0.001) in all data sets. The value of PVPI values followed the same pattern (2.6; 2.7; 3.5; P < 0.001).
Severity categories of ARDS described by the Berlin definition have good predictive validity and may be associated with increased extravascular lung water and pulmonary vascular permeability. Trial registration: UMIN-CTR ID UMIN000003627.
The selection of mechanical ventilator settings that ensure adequate oxygenation and carbon dioxide clearance while minimizing the risk of ventilator-associated lung injury (VALI) is a significant challenge for intensive-care clinicians. Current guidelines are largely based on previous experience combined with recommendations from a limited number of in vivo studies whose data are typically more applicable to populations than to individuals suffering from particular diseases of the lung. By combining validated computational models of pulmonary pathophysiology with global optimization algorithms, we generate in silico experiments to examine current practice and uncover optimal combinations of ventilator settings for individual patient and disease states. Formulating the problem as a multiobjective, multivariable constrained optimization problem, we compute settings of tidal volume, ventilation rate, inspiratory/expiratory ratio, positive end-expiratory pressure and inspired fraction of oxygen that optimally manage the trade-offs between ensuring adequate oxygenation and carbon dioxide clearance and minimizing the risk of VALI for different pulmonary disease scenarios.
The explosion of image data on the Internet has the potential to foster more sophisticated and robust models and algorithms to index, retrieve, organize and interact with images and multimedia data. But exactly how such data can be harnessed and organized remains a critical problem. We introduce here a new database called ldquoImageNetrdquo, a large-scale ontology of images built upon the backbone of the WordNet structure. ImageNet aims to populate the majority of the 80,000 synsets of WordNet with an average of 500-1000 clean and full resolution images. This will result in tens of millions of annotated images organized by the semantic hierarchy of WordNet. This paper offers a detailed analysis of ImageNet in its current state: 12 subtrees with 5247 synsets and 3.2 million images in total. We show that ImageNet is much larger in scale and diversity and much more accurate than the current image datasets. Constructing such a large-scale database is a challenging task. We describe the data collection scheme with Amazon Mechanical Turk. Lastly, we illustrate the usefulness of ImageNet through three simple applications in object recognition, image classification and automatic object clustering. We hope that the scale, accuracy, diversity and hierarchical structure of ImageNet can offer unparalleled opportunities to researchers in the computer vision community and beyond.
While our understanding of the pathogenesis and management of acute respiratory distress syndrome (ARDS) has improved over the past decade, estimates of its incidence have been controversial. The goal of this study was to examine ARDS incidence and outcome under current lung protective ventilatory support practices before and after the diagnosis of ARDS.
This was a 1-year prospective, multicenter, observational study in 13 geographical areas of Spain (serving a population of 3.55 million at least 18 years of age) between November 2008 and October 2009. Subjects comprised all consecutive patients meeting American-European Consensus Criteria for ARDS. Data on ventilatory management, gas exchange, hemodynamics, and organ dysfunction were collected.
A total of 255 mechanically ventilated patients fulfilled the ARDS definition, representing an incidence of 7.2/100,000 population/year. Pneumonia and sepsis were the most common causes of ARDS. At the time of meeting ARDS criteria, mean PaO(2)/FiO(2) was 114 ± 40 mmHg, mean tidal volume was 7.2 ± 1.1 ml/kg predicted body weight, mean plateau pressure was 26 ± 5 cmH(2)O, and mean positive end-expiratory pressure (PEEP) was 9.3 ± 2.4 cmH(2)O. Overall ARDS intensive care unit (ICU) and hospital mortality was 42.7% (95%CI 37.7-47.8) and 47.8% (95%CI 42.8-53.0), respectively.
This is the first study to prospectively estimate the ARDS incidence during the routine application of lung protective ventilation. Our findings support previous estimates in Europe and are an order of magnitude lower than those reported in the USA and Australia. Despite use of lung protective ventilation, overall ICU and hospital mortality of ARDS patients is still higher than 40%.
We aimed to validate the mathematical validity and accuracy of the respiratory components of the Nottingham Physiology Simulator (NPS), a computer simulation of physiological models. Subsequently, we aimed to assess the accuracy of the NPS in predicting the effects of a change in mechanical ventilation on patient arterial blood-gas tensions. The NPS was supplied with the following measured or calculated values from patients receiving intensive therapy: pulmonary shunt and physiological deadspace fractions, oxygen consumption, respiratory quotient, cardiac output, inspired oxygen fraction, expired minute volume, haemoglobin concentration, temperature and arterial base excess. Values calculated by the NPS for arterial oxygen tension and saturation (PaO2 and SaO2), mixed venous oxygen tension and saturation (PvO2 and SvO2), arterial and mixed venous carbon dioxide tension (PaCO2 and PvCO2) and arterial pH were accurate compared with measured values. Subsequently, arterial gas responses to changes in minute volume of FiO2 were measured in 31 patients and were compared with the NPS prediction for each response. The 95% limits of agreement in predicting the magnitude of change were: arterial oxygen tension -2.07 to 2.47 kPa; PaCO2 -0.33 to 0.67 kPa; and pH -0.023 to 0.033. This investigation has validated respiratory components of the NPS. We recommend the NPS as a clinical tool for predicting the effects of alterations in mechanical ventilation in stable patients in the intensive care unit.
This paper presents the methodology used in patient-specific calibration of a novel highly integrated model of the cardiovascular and pulmonary pathophysiology associated with Acute Respiratory Distress Syndrome (ARDS). We focus on data from previously published clinical trials on the static and dynamic cardio-pulmonary responses of three ARDS patients to changes in ventilator settings. From this data, the parameters of the integrated model were identified using an optimization-based methodology in multiple stages. Computational simulations confirm that the resulting model outputs accurately reproduce the available clinical data. Our results open up the possibility of creating in silico a biobank of virtual ARDS patients that could be used to evaluate current, and investigate novel, therapeutic strategies.
The respiratory-distress syndrome in 12 patients was manifested by acute onset of tachypnœa, hypoxæmia, and loss of compliance after a variety of stimuli; the syndrome did not respond to usual and ordinary methods of respiratory therapy. The clinical and pathological features closely resembled those seen in infants with respiratory distress and to conditions in congestive atelectasis and postperfusion lung. The theoretical relationship of this syndrome to alveolar surface active agent is postulated. Positive end-expiratory pressure was most helpful in combating atelectasis and hypoxæmia. Corticosteroids appeared to have value in the treatment of patients with fat-embolism and possibly viral pneumonia.