The 2021 SIIM-FISABIO-RSNA Machine Learning COVID-19 Challenge: Annotation and Standard Exam Classification of COVID-19 Chest Radiographs.

We describe the curation, annotation methodology, and characteristics of the dataset used in an artificial intelligence challenge for detection and localization of COVID-19 on chest radiographs. The chest radiographs were annotated by an international group of radiologists into four mutually exclusive categories, including "typical," "indeterminate," and "atypical appearance" for COVID-19, or "negative for pneumonia," adapted from previously published guidelines, and bounding boxes were placed on airspace opacities. This dataset and respective annotations are available to researchers for academic and noncommercial use.

Automated examination notification of emergency department images in a picture archiving and communication system.

This study compares the timeliness of radiology interpretation of Emergency Department (ED) imaging examinations in a picture archiving and communication system (PACS) before and after implementation of an automated paging system for notification of image availability. An alphanumeric pager for each radiology subspecialty (chest, pediatrics, bone, neuroradiology, and body) was used to alert the responsible radiologist that an ED imaging examination is available to be viewed on the PACS. The paging system was programmed to trigger off of the PACS database when an image is received on the appropriate radiology display station. The pager message includes the radiology accession number and examination description (such as chest, two-view, or c-spine, etc). The PACS paging tool performance was assessed by calculating the time elapsed, for each ED imaging examination, from the Time Imaged to the Time of Interpretation, where the Time Imaged is the actual image completion time measured at the imaging modality, and the Time Interpreted is the time a radiology interpretation is rendered to the ED, and is measured from the Radiology-to-ED fax time stamp. These measures were analyzed pre- and post-paging system implementation to determine any impact of the automated notification tool on radiology service turnaround time. Results show an improved radiology response time from image completion to interpretation rendered to ED clinicians, down from hour(s) to minutes, with the automated paging examination notification system. Examinations are read by the appropriate radiology specialty section in a more timely fashion, and fewer cases go unread by radiology.

Finding the optimal picture archiving and communication system (PACS) architecture: a comparison of three PACS designs.

PURPOSE: At present, there are two basic picture archiving and communication system (PACS) architectures: centralized with a central cache and controller, and distributed with a distributed cache and central controller. A third architecture proposed here is an autonomous one with a distributed cache and no controller. This report will investigate the performance (as measured be central processing unit [CPU] and network load, scalability, and examination retrieval and display latency) of these three types. METHODS: The distributed PACS architecture will be simulated using an IMPAX R3.5 (AGFA, Ridgefield Park, NJ) PACS, while the centralized design will be simulated using an IMPAX R4 (AGFA) PACS. The autonomous system will be designed and implemented in-house. The autonomous system consists of two types of entities: basic components such as acquisition gateways, display stations, and long-term archives, and registry servers, which store global state information about the individual PACS components. The key feature of the autonomous system will be the replacement of the central PACS controller by the registry servers. In this scenario the registry servers monitor the interactions between the components, but do not directly govern them. Instead each component will contain the application logic it requires and will use the state information from the registry servers to take the appropriate action, such as routing images, prefetching studies, and expiring images from near line cache. In addition the routing of examinations will be optimized to reduce the duplication of image data. Display stations will be categorized by specialty (neuroradiology, pediatrics, chest, etc) and will retrieve studies for display on demand from intermediate servers dedicated to the corresponding specialty. Studies will be routed only to the intermediate servers and not to display stations. RESULTS: By distributing the application logic, an autonomous PACS architecture can provide increased fault tolerance and therefore increased uptime. In addition, the lack of a central controller and the use of intermediate servers improve the scalability of the system, as well as reduce CPU and network loads.

What Digital Imaging and Communication in Medicine (DICOM) could look like in common object request broker (CORBA) and extensible markup language (XML).

Common object request broker architecture (CORBA) is a method for invoking distributed objects across a network. There has been some activity in applying this software technology to Digital Imaging and Communications in Medicine (DICOM), but no documented demonstration of how this would actually work. We report a CORBA demonstration that is functionally equivalent and in some ways superior to the DICOM communication protocol. In addition, in and outside of medicine, there is great interest in the use of extensible markup language (XML) to provide interoperation between databases. An example implementation of the DICOM data structure in XML will also be demonstrated. Using Visibroker ORB from Inprise (Scotts Valley, CA), a test bed was developed to simulate the principle DICOM operations: store, query, and retrieve (SQR). SQR is the most common interaction between a modality device application entity (AE) such as a computed tomography (CT) scanner, and a storage component, as well as between a storage component and a workstation. The storage of a CT study by invoking one of several storage objects residing on a network was simulated and demonstrated. In addition, XML database descriptors were used to facilitate the transfer of DICOM header information between independent databases. CORBA is demonstrated to have great potential for the next version of DICOM. It can provide redundant protection against single points of failure. XML appears to be an excellent method of providing interaction between separate databases managing the DICOM information object model, and may therefore eliminate the common use of proprietary client-server databases in commercial implementations of picture archiving and communication systems (PACS).

A hierarchical storage management (HSM) scheme for cost-effective on-line archival using lossy compression.

A hierarchical storage management (HSM) scheme for cost-effective on-line archival of image data using lossy compression is described. This HSM scheme also provides an off-site tape backup mechanism and disaster recovery. The full-resolution image data are viewed originally for primary diagnosis, then losslessly compressed and sent off site to a tape backup archive. In addition, the original data are wavelet lossy compressed (at approximately 25:1 for computed radiography, 10:1 for computed tomography, and 5:1 for magnetic resonance) and stored on a large RAID device for maximum cost-effective, on-line storage and immediate retrieval of images for review and comparison. This HSM scheme provides a solution to 4 problems in image archiving, namely cost-effective on-line storage, disaster recovery of data, off-site tape backup for the legal record, and maximum intermediate storage and retrieval through the use of on-site lossy compression.

Computers in imaging and health care: now and in the future.

Early picture archiving and communication systems (PACS) were characterized by the use of very expensive hardware devices, cumbersome display stations, duplication of database content, lack of interfaces to other clinical information systems, and immaturity in their understanding of the folder manager concepts and workflow reengineering. They were implemented historically at large academic medical centers by biomedical engineers and imaging informaticists. PACS were nonstandard, home-grown projects with mixed clinical acceptance. However, they clearly showed the great potential for PACS and filmless medical imaging. Filmless radiology is a reality today. The advent of efficient softcopy display of images provides a means for dealing with the ever-increasing number of studies and number of images per study. Computer power has increased, and archival storage cost has decreased to the extent that the economics of PACS is justifiable with respect to film. Network bandwidths have increased to allow large studies of many megabytes to arrive at display stations within seconds of examination completion. PACS vendors have recognized the need for efficient workflow and have built systems with intelligence in the management of patient data. Close integration with the hospital information system (HIS)-radiology information system (RIS) is critical for system functionality. Successful implementation of PACS requires integration or interoperation with hospital and radiology information systems. Besides the economic advantages, secure rapid access to all clinical information on patients, including imaging studies, anytime and anywhere, enhances the quality of patient care, although it is difficult to quantify. Medical image management systems are maturing, providing access outside of the radiology department to images and clinical information throughout the hospital or the enterprise via the Internet. Small and medium-sized community hospitals, private practices, and outpatient centers in rural areas will begin realizing the benefits of PACS already realized by the large tertiary care academic medical centers and research institutions. Hand-held devices and the Worldwide Web are going to change the way people communicate and do business. The impact on health care will be huge, including radiology. Computer-aided diagnosis, decision support tools, virtual imaging, and guidance systems will transform our practice as value-added applications utilizing the technologies pushed by PACS development efforts. Outcomes data and the electronic medical record (EMR) will drive our interactions with referring physicians and we expect the radiologist to become the informaticist, a new version of the medical management consultant.

Relevant priors prefetching algorithm performance for a picture archiving and communication system.

Proper prefetching of relevant prior examinations from a picture archiving and communication system (PACS) archive, when a patient is scheduled for a new imaging study, and sending the historic images to the display station where the new examination is expected to be routed and subsequently read out, can greatly facilitate interpretation and review, as well as enhance radiology departmental workflow and PACS performance. In practice, it has proven extremely difficult to implement an automatic prefetch as successful as the experienced fileroom clerk. An algorithm based on defined metagroup categories for examination type mnemonics has been designed and implemented as one possible solution to the prefetch problem. The metagroups such as gastrointestinal (GI) tract, abdomen, chest, etc, can represent, in a small number of categories, the several hundreds of examination types performed by a typical radiology department. These metagroups can be defined in a table of examination mnemonics that maps a particular mnemonic to a metagroup or groups, and vice versa. This table is used to effect the prefetch rules of relevance. A given examination may relate to several prefetch categories, and preferences are easily configurable for a particular site. The prefetch algorithm metatable was implemented in database structured query language (SQL) using a many-to-many fetch category strategy. Algorithm performance was measured by analyzing the appropriateness of the priors fetched based on the examination type of the current study. Fetched relevant priors, missed relevant priors, fetched priors that were not relevant to the current examination, and priors not fetched that were not relevant were used to calculate sensitivity and specificity for the prefetch method. The time required for real-time requesting of priors not previously prefetched was also measured. The sensitivity of the prefetch algorithm was determined to be 98.3% and the specificity 100%. Time required for on-demand requesting of priors was 9.5 minutes on average, although this time varied based on age of the prior examination and on the time of day and database traffic. A prefetch algorithm based on metatable examination mnemonic categories can pull the most appropriate relevant priors, reduce the number of missed relevant priors, and therefore reduce the time involved for the manual task of on-demand requests of priors. Network and database traffic can be reduced as well by decreasing the number of priors selected from the archive and subsequently transmitted to the display stations, through elimination of transactions on examinations not relevant to the current study.

Simulation of disaster recovery of a picture archiving and communications system using off-site hierarchal storage management.

The purpose of this communication is to report on the testing of the disaster recovery capability of our hierarchical storage management (HSM) system. Disaster recovery implementation is a requirement of every mission-critical information technology project. Picture archiving and communications systems (PACS) certainly falls into this category, even though the counterpart, conventional film archive, has no protection against fire, for example. We have implemented a method for hierarchical storage with wavelet technology that maximizes on-site case storage (using lossy compression), retains bit-preserved image data for legal purposes, provides an off-site backup (lossless bit-preserving wavelet transform), and provides for disaster recovery. Recovery from a natural (earthquake and subsequent fire) or technical (system crash and data loss) disaster was simulated by attempting to restore from the off-site image and database backup to clean core PACS components. The only existing loaded software was the operating system. The database application was reloaded locally, and then the database contents and image store were loaded from the off-site component of the HSM system. The following measurements were analyzed: (1) the ability to recover all data; (2) the integrity of the recovered database and image data; (3) the time to recover the database relative to the number of studies and age of the archive, as well as bandwidth between the local and remote site; and (4) the time to recover image data relative to compression ratio, number of studies, number of images, and time depth of the archive. This HSM system, which maximizes on-site storage, maintains a legal record, and provides off-site backup, also facilitates disaster recovery for a PACS.

PACS databases and enrichment of the folder manager concept.

Current challenges facing picture archiving and communication systems (PACS) center around database design and functionality. Workflow issues and folder manager concepts such as autorouting, prefetching, hanging protocols, and hierarchical storage management are driven by a properly designed database that ultimately directly impacts the clinical utility of a PACS. The key issues in PACS database design that enable radiologist-friendly, cost-effective, and data-secure systems will be discussed, including database difficulties of the DICOM standard, HIS/RIS/PACS (hospital information system/radiology information system) connectivity, and database issues in data acquisition, data dissemination, and data display.

Finding-specific display presets for computed radiography soft-copy reading.

Much work has been done to optimize the display of cross-sectional modality imaging examinations for soft-copy reading (i.e., window/level tissue presets, and format presentations such as tile and stack modes, four-on-one, nine-on-one, etc). Less attention has been paid to the display of digital forms of the conventional projection x-ray. The purpose of this study is to assess the utility of providing presets for computed radiography (CR) soft-copy display, based not on the window/level settings, but on processing applied to the image optimized for visualization of specific findings, pathologies, etc (i.e., pneumothorax, tumor, tube location). It is felt that digital display of CR images based on finding-specific processing presets has the potential to: speed reading of digital projection x-ray examinations on soft copy; improve diagnostic efficacy; standardize display across examination type, clinical scenario, important key findings, and significant negatives; facilitate image comparison; and improve confidence in and acceptance of soft-copy reading. Clinical chest images are acquired using an Agfa-Gevaert (Mortsel, Belgium) ADC 70 CR scanner and Fuji (Stamford, CT) 9000 and AC2 CR scanners. Those demonstrating pertinent findings are transferred over the clinical picture archiving and communications system (PACS) network to a research image processing station (Agfa PS5000), where the optimal image-processing settings per finding, pathologic category, etc, are developed in conjunction with a thoracic radiologist, by manipulating the multiscale image contrast amplification (Agfa MUSICA) algorithm parameters. Soft-copy display of images processed with finding-specific settings are compared with the standard default image presentation for 50 cases of each category. Comparison is scored using a 5-point scale with the positive scale denoting the standard presentation is preferred over the finding-specific processing, the negative scale denoting the finding-specific processing is preferred over the standard presentation, and zero denoting no difference. Processing settings have been developed for several findings including pneumothorax and lung nodules, and clinical cases are currently being collected in preparation for formal clinical trials. Preliminary results indicate a preference for the optimized-processing presentation of images over the standard default, particularly by inexperienced radiology residents and referring clinicians.

Anatomy of picture archiving and Communications systems: nuts and bolts--image acquisition: getting digital images from imaging modalities.

Digital acquisition of data from the various imaging modalities for input to a picture archiving and communication system (PACS) is discussed. Essential features for successful clinical implementation including Digital Imaging and Communications in Medicine (DICOM) compliance, radiology information system (RIS)/hospital information system (HIS) interfacing, and workflow integration are detailed. Image acquisition from the inherently digital cross-sectional modalities are described, as well as digital acquisition of the conventional projection x-ray using computed radiography (CR), direct digital radiography (DDR), and film digitizers.

Continuing quality improvement procedures for a clinical PACS.

The University of California at San Francisco (USCF) Department of Radiology currently has a clinically operational picture archiving and communication system (PACS) that is thirty-five percent filmless, with the goal of becoming seventy-five percent filmless within the year. The design and implementation of the clinical PACS has been a collaborative effort between an academic research laboratory and a commercial vendor partner. Images are digitally acquired from three computed radiography (CR) scanners, five computed tomography (CT) scanners, five magnetic resonance (MR) imagers, three digital fluoroscopic rooms, an ultrasound mini-PACS and a nuclear medicine mini-PACS. The DICOM (Digital Imaging and Communications in Medicine) standard communications protocol and image format is adhered to throughout the PACS. Images are archived in hierarchical staged fashion, on a RAID (redundant array of inexpensive disks) and on magneto-optical disk jukeboxes. The clinical PACS uses an object-oriented Oracle SQL (systems query language) database, and interfaces to the Radiology Information System using the HL7 (Health Languages 7) standard. Components are networked using a combination of switched and fast ethernet, and ATM (asynchronous transfer mode), all over fiber optics. The wide area network links six UCSF sites in San Francisco. A combination of high and medium resolution dual-monitor display stations have been placed throughout the Department of Radiology, the Emergency Department (ED) and Intensive Care Units (ICU). A continuing quality improvement (CQI) committee has been formed to facilitate the PACS installation and training, workflow modifications, quality assurance and clinical acceptance. This committee includes radiologists at all levels (resident, fellow, attending), radiology technologists, film library personnel, ED and ICU clinician end-users, and PACS team members. The CQI committee has proved vital in the creation of new management procedures, providing a means for user feedback and education, and contributing to the overall acceptance of, and user satisfaction with the system. Well developed CQI procedures have been essential to the successful clinical operation of the PACS as UCSF Radiology moves toward a filmless department.

Optimization of a low-cost truly preemptive multitasking PC diagnostic workstation.

The Windows 95/NT operating systems (Microsoft Corp, Redmond, WA) currently provide the only low-cost truly preemptive multitasking environment and as such become an attractive diagnostic workstation platform. The purpose of this project is to test and optimize display station graphical user interface (GUI) actions previously designed on the pseudomultitasking Macintosh (Apple Computer, Cupertino, CA) platform, and image data transmission using time slicing/ dynamic prioritization assignment capabilities of the new Windows platform. A diagnostic workstation in the clinical environment must process two categories of events: user interaction with the GUI through keyboard/mouse input, and transmission of incoming data files. These processes contend for central processing units (CPU) time resulting in GUI "lockout" during image transmission or delay in transmission until GUI "quiet time." WinSockets and the Transmission Control Protocol/Internet Protocal (TCP/IP) communication protocol software (Microsoft) are implemented using dynamic priority timeslicing to ensure that GUI delays at the time of Digital Imaging and Communications in Medicine (DICOM) file transfer do not exceed 1/10 second. Assignment of thread priority does not translate into an absolute fixed percentage of CPU time. Therefore, the relationship between dynamic priority assignment by the processor, and the GUI and communication application threads will be more fully investigated to optimize CPU resource allocation. These issues will be tested using 10 MB/sec Ethernet and 100 MB/sec fast and wide Ethernet transmission. Preliminary results of typical clinical files (10 to 30 MB) over Ethernet show no visually perceptible interruption of the GUI, suggesting that the new Windows PC platform may be a viable diagnostic workstation option.

Viewport: an object-oriented approach to integrate workstation software for tile and stack mode display.

Diagnostic workstation design has migrated towards display presentation in one of two modes: tiled images or stacked images. It is our impression that the workstation setup or configuration in each of these two modes is rather distinct. We sought to establish a commonality to simplify software design, and to enable a single descriptor method to facilitate folder manager development of "hanging" protocols. All current workstation designs use a combination of "off-screen" and "on-screen" memory whether or not they use a dedicated display subsystem, or merely a video board. Most diagnostic workstations also have two or more monitors. Our central concept is that of a "logical" viewport that can be smaller than, the same size as, or larger than a single monitor. Each port "views" an image data sequence loaded into offscreen memory. Each viewport can display one or more images in sequence in a one-on-one or traditionally tiled presentation. Viewports can be assigned to the available monitor "real estate" in any manner that fits. For example, a single sequence computed tomography (CT) study could be displayed across all monitors in a tiled appearance by assigning a single large viewport to the monitors. At the other extreme, a multisequence magnetic resonance (MR) study could be compared with a similar previous study by assigning four viewports to each monitor, single image display per viewport, and assigning four of the sequences of the current study to the left monitor viewports, and four of the earlier study to the right monitor viewports. Ergonomic controls activate scrolling through the off-screen image sequence data. Workstation folder manager hanging protocols could then specify viewports, number of images per viewport, and the automatic assignment of appropriately named sequences of current and previous studies to the viewports on a radiologist-specific basis. Furthermore, software development is simplified by common base objects and methods of the tile and stack modes. Prototype workstation display software and folder manager protocol implementation will be described and demonstrated.

Artifacts in chest radiographs with a third-generation computed radiography system.

Photostimulable phosphor computed radiography (CR) is a developing and increasingly widespread technology. The purpose of this pictorial essay is to familiarize readers with the appearance and cause of image artifacts that can occur in a third-generation computed radiographic system. Artifacts are described that relate to imaging plates, image readers, image processing, and film processing.

Multimedia in the radiology environment: current concept.

Multimedia has different meanings according to its context. Here, multimedia in the radiology environment is defined as the integration of multiple radiology and medical information systems to facilitate the practice of radiology. These information systems include the hospital information system, radiology information system, picture archiving and communication systems, voice reporting, library information systems, and electronic mail and file systems. The concept of multimedia within the context of integration of these database systems will be presented. An example is given on how to access these information systems by a radiologist's desktop personal computer.