Dataset Information

MIMIC-CXR, a de-identified publicly available database of chest radiographs with free-text reports.

ABSTRACT: Chest radiography is an extremely powerful imaging modality, allowing for a detailed inspection of a patient's chest, but requires specialized training for proper interpretation. With the advent of high performance general purpose computer vision algorithms, the accurate automated analysis of chest radiographs is becoming increasingly of interest to researchers. Here we describe MIMIC-CXR, a large dataset of 227,835 imaging studies for 65,379 patients presenting to the Beth Israel Deaconess Medical Center Emergency Department between 2011-2016. Each imaging study can contain one or more images, usually a frontal view and a lateral view. A total of 377,110 images are available in the dataset. Studies are made available with a semi-structured free-text radiology report that describes the radiological findings of the images, written by a practicing radiologist contemporaneously during routine clinical care. All images and reports have been de-identified to protect patient privacy. The dataset is made freely available to facilitate and encourage a wide range of research in computer vision, natural language processing, and clinical data mining.

SUBMITTER: Johnson AEW

PROVIDER: S-EPMC6908718 | biostudies-literature | 2019 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

MIMIC-CXR, a de-identified publicly available database of chest radiographs with free-text reports.

Johnson Alistair E W AEW Pollard Tom J TJ Berkowitz Seth J SJ Greenbaum Nathaniel R NR Lungren Matthew P MP Deng Chih-Ying CY Mark Roger G RG Horng Steven S

Scientific data 20191212 1

Chest radiography is an extremely powerful imaging modality, allowing for a detailed inspection of a patient's chest, but requires specialized training for proper interpretation. With the advent of high performance general purpose computer vision algorithms, the accurate automated analysis of chest radiographs is becoming increasingly of interest to researchers. Here we describe MIMIC-CXR, a large dataset of 227,835 imaging studies for 65,379 patients presenting to the Beth Israel Deaconess Medi ...[more]

PMID: 31831740

Similar Datasets

Project description:BackgroundWorkplace violence against medical staff in China is a widespread problem that has negative impacts on medical service delivery. The study aimed to contribute to the prevention of workplace violence against medical staff in China by identifying patterns of workplace violence, key risk factors, and the interplay of risk factors that result in workplace violence.MethodsNinety-seven publicly reported Chinese healthcare violent incidents from late 2013 to 2017 were retrospectively collected from the internet and analysed using content analysis. A modified socio-ecological model guided analysis of the violent incidents focusing on risk.ResultsPhysical violence, yinao, or a combination of physical and verbal violence were the typical forms of violence reported. The findings identified risk at all levels. Individual level risk factors included service users' unreasonable expectations, limited health literacy, mistrust towards medical staff, and inadequacy of medical staff's communication during the medical encounter. Organisational level risk factors under the purview of hospital management included problems with job design and service provision system, inadequacies with environmental design, security measures, and violence response mechanisms within hospitals. Societal level risk factors included lack of established medical dispute-handling mechanisms, problems in legislation, lack of trust and basic health literacy among service users. Situational level risks were contingent on risk factors on the other levels: individual, organisational, and societal.ConclusionsInterventions at individual, situational, organisational, and societal levels are needed to systematically address workplace violence against medical staff in China. Specifically, improving health literacy can empower patients, increase trust in medical staff and lead to more positive user experiences. Organizational-level interventions include improving human resource management and service delivery systems, as well as providing training on de-escalation and violence response for medical staff. Addressing risks at the societal level through legislative changes and health reforms is also necessary to ensure medical staff safety and improve medical care in China.

Project description:BackgroundThe air traffic management (ATM) system has historically coped with a global increase in traffic demand ultimately leading to increased operational complexity. When dealing with the impact of this increasing complexity on system safety it is crucial to automatically analyse the losses of separation (LoSs) using tools able to extract meaningful and actionable information from safety reports. Current research in this field mainly exploits natural language processing (NLP) to categorise the reports,with the limitations that the considered categories need to be manually annotated by experts and that general taxonomies are seldom exploited.MethodsTo address the current gaps,authors propose to perform exploratory data analysis on safety reports combining state-of-the-art techniques like topic modelling and clustering and then to develop an algorithm able to extract the Toolkit for ATM Occurrence Investigation (TOKAI) taxonomy factors from the free-text safety reports based on syntactic analysis. TOKAI is a tool for investigation developed by EUROCONTROL and its taxonomy is intended to become a standard and harmonised approach to future investigations.ResultsLeveraging on the LoS events reported in the public databases of the Comisión de Estudio y Análisis de Notificaciones de Incidentes de Tránsito Aéreo and the United Kingdom Airprox Board,authors show how their proposal is able to automatically extract meaningful and actionable information from safety reports,other than to classify their content according to the TOKAI taxonomy. The quality of the approach is also indirectly validated by checking the connection between the identified factors and the main contributor of the incidents.ConclusionsAuthors' results are a promising first step toward the full automation of a general analysis of LoS reports supported by results on real-world data coming from two different sources. In the future,authors' proposal could be extended to other taxonomies or tailored to identify factors to be included in the safety taxonomies.

Project description:BACKGROUND: Access to unpublished clinical study reports (CSRs) is currently being discussed as a means to allow unbiased evaluation of clinical research. The Institute for Quality and Efficiency in Health Care (IQWiG) routinely requests CSRs from manufacturers for its drug assessments. Our objective was to determine the information gain from CSRs compared to publicly available sources (journal publications and registry reports) for patient-relevant outcomes included in IQWiG health technology assessments (HTAs) of drugs. METHODS AND FINDINGS: We used a sample of 101 trials with full CSRs received for 16 HTAs of drugs completed by IQWiG between 15 January 2006 and 14 February 2011, and analyzed the CSRs and the publicly available sources of these trials. For each document type we assessed the completeness of information on all patient-relevant outcomes included in the HTAs (benefit outcomes, e.g., mortality, symptoms, and health-related quality of life; harm outcomes, e.g., adverse events). We dichotomized the outcomes as "completely reported" or "incompletely reported." For each document type, we calculated the proportion of outcomes with complete information per outcome category and overall. We analyzed 101 trials with CSRs; 86 had at least one publicly available source, 65 at least one journal publication, and 50 a registry report. The trials included 1,080 patient-relevant outcomes. The CSRs provided complete information on a considerably higher proportion of outcomes (86%) than the combined publicly available sources (39%). With the exception of health-related quality of life (57%), CSRs provided complete information on 78% to 100% of the various benefit outcomes (combined publicly available sources: 20% to 53%). CSRs also provided considerably more information on harms. The differences in completeness of information for patient-relevant outcomes between CSRs and journal publications or registry reports (or a combination of both) were statistically significant for all types of outcomes. The main limitation of our study is that our sample is not representative because only CSRs provided voluntarily by pharmaceutical companies upon request could be assessed. In addition, the sample covered only a limited number of therapeutic areas and was restricted to randomized controlled trials investigating drugs. CONCLUSIONS: In contrast to CSRs, publicly available sources provide insufficient information on patient-relevant outcomes of clinical trials. CSRs should therefore be made publicly available. Please see later in the article for the Editors' Summary.

Dataset Information

MIMIC-CXR, a de-identified publicly available database of chest radiographs with free-text reports.

Publications

MIMIC-CXR, a de-identified publicly available database of chest radiographs with free-text reports.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets