Glossary of Enterprise Imaging Terms

Algorithm: A defined sequence of operations used by artificial intelligence (AI) systems or other computing solutions to process data and make decisions autonomously.

Adversarial machine learning (AML) is a field of study within ML that is focused on understanding and defending against adversarial attacks on ML models, such as manipulating input data to deceive or compromise the performance of ML models.

Annotation: Detailed clinical notes that are attached to healthcare assets, including diagrams, electronic health records (EHRs), and medical imaging files.

AI Application: The implementation of AI to perform domain-specific tasks, such as clinical decision support, robotic process automation, and radiological diagnostics.

API (Application Programming Interface): A set of protocols that enable different software applications to communicate. Common standards in healthcare imaging include DICOM, HL7, FHIR, and DICOMWeb RESTful APIs.

An Artificial Neural Network (ANN) is an AI model inspired by the human brain. It consists of interconnected nodes (neurons) that process and analyze complex data structures.

Automatic speech recognition (ASR) is a process for automatically converting speech into text. ASR is used in voice interfaces, virtual assistants, dictation systems, and voice-controlled devices, allowing hands-free interaction and improving accessibility for users.

Availability: According to the Society of Imaging Informatics (SIIM), availability measures when a system is fully available for the business functions designed. For a PACS, for example, this would mean support of DICOM image management, archival, and visualization. All essential system functions must be operating to qualify as available. (For example, image management and archive, but not visualization, would not be eligible for a PACS as “available” in the proper sense.)

Big Data: Enormous datasets that may be analyzed computationally to reveal patterns, trends, and associations, especially relating to human behavior and interactions. In enterprise imaging, big data refers to vast data generated within radiology departments. Four Vs characterize big data: volume (scale of data), velocity (speed of data processing), variety (different forms of data), and veracity (uncertainty of data). Potential applications in radiology include optimizing scan scheduling, developing personalized scanning protocols, providing decision support to radiologists, enhancing emergency reporting, and enabling virtual quality assurance.

Bias (Data Bias) is a systematic distortion in AI model training data that can affect fairness, accuracy, and transparency. Common biases in healthcare include sample bias, measurement bias, and confirmation bias.

Blockchain records transactions across multiple computers in a peer-to-peer network, ensuring data integrity and security. It functions as a decentralized digital ledger that duplicates and distributes transaction records across the entire network, making it highly resistant to tampering, hacking, or unauthorized alterations. While commonly associated with cryptocurrencies, blockchain technology extends to various applications, including supply chain management, healthcare data security, and digital identity verification.

Classification: A supervised machine learning task that categorizes data into predefined labels, commonly used in medical imaging to detect abnormalities

Cloud Computing: The delivery of computing services such as storage, processing, and networking over the internet, enabling healthcare organizations to access imaging data remotely and securely.

Healthcare enterprises use different methods to deploy cloud resources. There is a public cloud that shares resources and offers services to the public over the Internet, a private cloud that isn’t shared and offers services over a private internal network typically hosted on-premises, a hybrid cloud that shares services between public and private clouds depending on their purpose, and a community cloud that shares resources only between organizations, such as with government institutions.

Cloud PACS: A cloud-based Picture Archiving and Communication System (PACS) that allows medical images to be stored, accessed, and shared via cloud infrastructure rather than on-premises servers. Cloud PACS solutions provide scalable storage, remote access, enhanced collaboration among healthcare providers, and improved disaster recovery options.

Data Lake is a centralized repository that stores structured and unstructured data, allowing advanced analytics and AI model training. A data lake can include structured data from relational databases (rows and columns), semi-structured data (CSV, logs, XML, JSON), unstructured data (emails, documents, PDFs), and binary data (images, audio, video). A data lake can be established “on-premises” (within an organization’s data centers) or “in the cloud” (using cloud services from vendors such as Amazon, Microsoft, or Google).

Data Swamp is an unmanaged data lake that has become disorganized, reducing its accessibility and analytical value

Deep Learning: a type of machine learning that replicates the innate human ability to process data in abstract ways. The data must be processed through several ‘layers’ of meaning to conclude, as opposed to the relatively instinctive reasoning that a human can perform.

De-identification is the process of anonymizing patient data in compliance with regulations such as HIPAA, removing identifiable information from datasets. Common strategies include deleting or masking personal identifiers, such as personal names, and suppressing or generalizing quasi-identifiers, such as date of birth. The reverse process of using de-identified data to identify individuals is known as data re-identification.

DICOM: Digital Imaging and Communications in Medicine (DICOM) is the standard for communicating and managing medical imaging information and related data. DICOM is most commonly used for storing and transmitting medical images, enabling the integration of medical imaging devices such as scanners, servers, workstations, printers, network hardware, and PACS (picture archiving and communication systems) from multiple manufacturers.

DICOM Load Balancing: A process that optimizes the distribution of Digital Imaging and Communications in Medicine (DICOM) data traffic across multiple servers or network resources. DICOM load balancing involves intelligently distributing medical images across multiple servers or systems based on predefined rules and priorities. Timely access to imaging data enables healthcare providers to make more informed, data-driven clinical decisions.

DICOM Modality Worklist (MWL): The DICOM modality worklist service provides a list of imaging procedures scheduled for performance by an image acquisition device (sometimes referred to as a modality system). The items in the worklist include relevant details about the subject of the procedure (patient ID, name, sex, and age), the type of procedure (equipment type, procedure description, procedure code), and the procedure order (referring physician, accession number, reason for exam). An image acquisition device, such as a CT scanner, queries a service provider, such as a RIS or PACS, to get this information, which is then presented to the system operator and used by the imaging device to populate details in the image metadata.

DICOMweb is a term applied to the family of RESTful DICOM services defined for sending, retrieving, and querying medical images and related information. The intent is to provide a light-weight mobile device and web browser-friendly mechanism for accessing images used by developers who have minimal familiarity with the DICOM standard, and which uses consumer application-friendly mechanisms like HTTP, JSON, and media types (like “image/jpeg”) to the maximum extent possible.

A key feature of the WADO-RS services is the ability to retrieve entire studies and series without requiring repeated requests for individual instances.

DICOM Tags are metadata elements associated with each DICOM image object, containing essential information about the patient, study, and image acquisition parameters. These tags, represented by unique identifiers in the format (XXXX, XXXX), facilitate efficient management and analysis of medical imaging data by storing crucial details such as Patient ID, Study Date, Modality, and Slice Thickness.

Disaster Recovery (DR) is a set of best practices designed to prevent or minimize data loss and business disruption resulting from catastrophic events, including localized power outages, equipment failures, cyberattacks, civil emergencies, criminal or military attacks, and natural disasters like hurricanes, tornadoes, or earthquakes. VNA is essential to maintaining business continuity when such events occur. Whether on-premise or in the Cloud, a VNA safely stores files in various facilities and individual data centers. VNAs serve as backup DICOM image viewers. Additionally, organizations can rebuild a database after a disaster by using a VNA that leverages indexed metadata via a DICOM header.

Electronic Health Records (EHRs) systematically collect and digitally store patient and population health information. Healthcare providers can share these records across various settings through network-connected, enterprise-wide information systems or other information networks and exchanges. EHRs contain a wide range of data, including demographics, medical history, medications, allergies, immunization status, laboratory test results, radiology images, vital signs, personal statistics such as age and weight, and billing information.

Encounter-based Workflow: In this imaging workflow, healthcare providers perform imaging during a patient encounter without relying on a pre-existing order.

Enterprise imaging is a set of strategies, initiatives, and workflows implemented across a healthcare enterprise to consistently and optimally capture, index, manage, store, distribute, view, exchange, and analyze all medical imaging and multimedia content to enhance the electronic health record.

Enterprise imaging workflows are the processes or customizations involved in enabling interconnected, optimal, and efficient routing, storage, viewing, distributing, and exchanging of medical images and studies throughout healthcare organizations (enterprises). Scheduled Workflow (SWF) integrates the ordering, scheduling, imaging acquisition, storage, and viewing activities associated with radiology exams. Encounter-Based Imaging Workflow (EBIW) captures images acquired in an encounter between a patient and a healthcare provider.

Failback involves switching back to the original primary systems. Failback is the second stage of a two-part system for safeguarding information in a crisis mode during natural disasters or other events that can compromise an IT operation.

Failover is the ability to switch automatically and seamlessly to a reliable backup system. When a component or primary healthcare infrastructure system fails, either a standby operational mode or redundancy should achieve failover.

FHIR RESTful API: Fast Healthcare Interoperability Resources (FHIR, pronounced “fire”) is a standard describing data formats and elements (known as “resources”) and an application programming interface (API) for exchanging electronic health records (EHR). FHIR builds on previous data format standards from HL7 and is easier to implement because it uses a modern web-based suite of API technology, including an HTTP-based RESTful protocol.

F-score, also known as the F1-score, is a measure of a model’s accuracy that combines precision and recall into a single metric.

Generative AI – AI that creates new content or data, useful for synthetic image generation or augmentation.

Hanging Protocols provide a unique set of viewing instructions that determine the layout and display of medical images. They allow the user to view an image or study based on the modality type, number of images, comparison images, and other user-specified criteria. Predefined Hanging Protocols allow the viewer to immediately begin the interpretation process using optimal viewing settings for the specific image or exam type. Effective use of adequately configured hanging protocols alleviates the need to arrange and adjust images upon initial display.

Healthcare informatics integrates science and engineering principles to apply informatics methodologies to medical practices. It addresses complex challenges within the healthcare domain by leveraging computational techniques to optimize data management, clinical decision-making, and operational efficiency.

HIPAA Safe Harbor De-Identification is the process that removes specified identifiers from patient records, including those of the patient’s relatives, household members, and employers. Covered entities satisfy the requirements only when they have removed all specified identifiers and have no actual knowledge that the remaining information could identify the patient.

Once de-identified, protected health information (PHI) no longer falls under HIPAA regulations, allowing unrestricted use and disclosure. By definition, de-identified health information neither identifies nor provides a reasonable basis to identify a patient.

Healthcare Information System (HIS) is the core information system for patient management in the healthcare and radiology information systems.

HL7 is a set of international standards for transferring clinical and administrative data between software applications used by various healthcare providers. HL7 standards support managing, delivering, and evaluating clinical practice and health services.

Human-in-the-loop (HITL) is a model or system where human insight is integrated at various stages of the process to provide feedback, validation, and decision-making.

Interoperability is the ability of different information systems, devices, and applications (systems) to access, exchange, integrate, and cooperatively use data in a coordinated manner within and across organizational, regional, and national boundaries to provide timely and seamless portability of information and optimize the health of individuals and populations globally.

Health data exchange architectures, application interfaces, and standards enable data to be accessed and shared appropriately and securely across the complete spectrum of care within all applicable settings and with relevant stakeholders, including the individual.

Image Lifecycle Management manages medical images and data storage while offering tools to migrate content between repositories as organizations adopt new modalities and retire older ones. It optimizes the lifespan of medical images and data, ensuring organized and accessible storage. Centralizing data in a VNA streamlines management and minimizes the complexities of handling multiple storage systems.

Image Object Change Management maintains data integrity across systems by enabling one system to communicate updates to others, managing copies of the same imaging objects within their local environments. It facilitates critical actions such as study deletions, image removals, and study merges. In a VNA, Image Object Change Management ensures consistent data across interconnected systems by defining protocols for transmitting local modifications, including study consolidations, deletions, and expungements of medical images.

Image segmentation: a task performed by AI dividing up a digital image into regions corresponding to the image contents, such as visually identifying the different parts of a car.

Information technology (IT) uses computers to store or retrieve data and information, typically used within business operations as opposed to personal or entertainment technologies. An information technology system (IT system) is generally an information system, a communications system, or, more precisely, a computer system – including all hardware, software, and peripheral equipment – operated by a limited group of IT users.

Integrating the Healthcare Enterprise (IHE): This is an initiative by healthcare professionals and the industry to improve the way computer systems in healthcare share information. It brings together users and developers of healthcare information technology (HIT) in an annually recurring four-step process.

Intelligent Routing enables complex image routing that PACS systems cannot support. It handles tasks such as routing special procedures, rerouting images for study segmentation, and managing unique exception workflows. The system can trigger routing based on various criteria, including standard or private DICOM tags, specific dates or times, HL7 fields, or groups of fields. It also caches data for future delivery and performs checks to ensure the complete transfer of imaging studies.

Machine learning: a process whereby a machine learns and changes without human prompting, based on the data it is fed.’ Over time, it will recognize patterns and adapt to predict outcomes based on the fed data.

Masking uses software such as Photoshop to hide and reveal portions of an image. It is a non-destructive image editing process and a post-processing task. For example, in radiography, masking enables the comparative study of small areas of both breasts in mammograms.

Natural language processing (NLP) is a subfield of linguistics, computer science, and artificial intelligence concerned with the interactions between computers and human language, mainly how to program computers to process and analyze large amounts of natural language data. The goal is a computer capable of “understanding” the contents of documents, including the contextual nuances of the language within them. The technology can then accurately extract information and insights from the documents and categorize and organize them.

Non-DICOM Content consists of file formats such as PDF, XML, and images produced by specialized equipment and modalities. VNAs store both non-DICOM content and DICOM images, ensuring seamless integration. For example, they can store scanned documents as PDFs, which PACS systems then convert into DICOM PDF objects with the necessary metadata for easy identification and management. The ability to manage all types of non-DICOM data alongside DICOM images makes VNAs a valuable solution for comprehensive healthcare data management.

Optical character recognition (OCR): a technology that converts different types of documents, such as scanned paper documents, PDF files, or images captured by a digital camera, into editable and searchable data. OCR is the electronic or mechanical conversion of images of typed, handwritten, or printed text into machine-encoded text, whether from a scanned document, a photo of a document, a scene photo (for example, the text on signs and billboards in a landscape photo) or from subtitle text superimposed on an image (for example: from a television broadcast).

On-premises or on-prem software is software installed and runs on computers on the premises of the person or organization using the software rather than at a remote facility such as a server farm or cloud. On-premises software is sometimes referred to as “shrinkwrap” software, and off-premises software is commonly called “software as a service” (“SaaS”) or “cloud computing.”

Order-based Workflow: In an order-based workflow, healthcare providers place an imaging order before performing the procedure, ensuring proper documentation and streamlined workflow management.

Overfitting: a term describing a problem sometimes encountered in supervised learning where a machine intelligence specializes in recognizing patterns in the curated data it has trained on, becoming unable to identify patterns in new data quickly.

PACS: A picture archiving and communication system (PACS) is a medical imaging technology that provides economical storage and convenient access to images from multiple modalities (source machine types). Electronic images and reports are transmitted digitally via PACS; this eliminates the need to manually file, retrieve, or transport film jackets, the folders used to store and protect X-ray film. The universal format for PACS image storage and transfer is DICOM. Non-image data, such as scanned documents, may be incorporated using consumer industry standard formats like PDF (Portable Document Format) once encapsulated in DICOM.

Protected Health Information (PHI): The HIPAA Privacy Rule provides federal protections for personal health information held by covered entities and gives patients various rights. The PHI Privacy Rule is balanced so that it permits the disclosure of personal health information needed for patient care and other essential purposes.

QIDO-RS: is a DICOMweb term applied to the family of RESTful DICOM services defined for sending, retrieving, and querying medical images and related information. QIDO-RS is a query based on ID for DICOM Objects (QIDO). It enables the search for studies, series, and instances by patient ID and receiving their unique identifiers for further usage (i.e., to retrieve their rendered representations).

Radiology uses imaging technology to diagnose and treat diseases. It consists of two main areas: diagnostic radiology, which focuses on identifying conditions, and interventional radiology, which involves minimally invasive imaging-guided treatments. Physicians specializing in radiology are known as radiologists.

Radiology Information System (RIS): A RIS is a networked software system for managing medical imagery and associated data. It is beneficial for tracking radiology imaging orders and billing information. It is often used with PACS and vendor-neutral archives (VNAs) to manage image archives, record-keeping, and billing.

Representational State Transfer (REST) is an API architecture that leverages existing protocols, most commonly HTTP for web APIs, eliminating the need for additional software or libraries during implementation. Its flexible design allows developers to handle multiple types of requests, return various data formats, and dynamically adapt to structural changes using hypermedia. REST is not tightly coupled to specific resources or methods, unlike other APIs, making it highly adaptable for diverse applications.

Recovery point objective (RPO) refers to the data you can lose in a disaster. RPO is the maximum tolerable length of time a computer, system, network, or application can be down after a failure or disaster occurs and continuously copies data to a remote data center so that an outage will not result in data loss. Some organizations determine that losing five minutes or one hour of data is acceptable.

Reinforcement learning (RL) is a machine learning paradigm where an agent learns to make decisions by interacting with an environment, receiving feedback in the form of rewards or penalties for its actions.

Relevant Priors refers to prior imaging studies that provide vital information and context to the radiologist when interpreting current research, which may lead to a possible diagnosis change. A “prefetch” step is standard in imaging workflows. Prefetching locates prior studies that may be relevant as context for reading the current research and makes them immediately accessible to the radiologist. Prefetching usually occurs in response to a preceding event, such as receiving an HL7 order or ADT message (ORM, ADT) or querying a DICOM Modality Worklist (DMWL) to identify upcoming exams. This proactive approach ensures that relevant prior studies are available before the radiologist begins reviewing the new exam.

When an HL7 message or DMWL is unavailable, a “postfetch” can be initiated upon receiving a DICOM object, such as the first image of a new exam or a DICOM Structured Report (SR) from a modality. However, prefetching is preferred because it minimizes delays and enhances workflow efficiency by retrieving prior studies in advance.

If prior exams are not readily available, radiologists may experience delays, leading to workflow disruptions and frustration. Prefetching remains effective even for locally stored prior exams, though retrieving them from long-term storage may take additional time.

Self-supervised learning (SSL) is a machine learning technique where a model learns to make predictions about certain parts of the input data using the other parts as context. Rather than relying on labeled datasets for supervisory signals, self-supervised models generate implicit labels from unstructured data.

Strong AI: a machine that thinks and communicates on the level of a human or higher. Currently theoretical, restricted to science fiction.

STOW-RS is a DICOMweb term applied to the family of RESTful DICOM services defined for sending, retrieving, and querying medical images and related information. STOW-RS is for storing (sending) DICOM PS3.10 files or separated metadata and bulk data.

Supervised learning is a form of machine learning where machine intelligence learns from annotated data samples to generate the desired output correctly. Its algorithm mathematically generalizes data patterns and, through close analysis, becomes (in theory) better at predicting patterns than humans.

Structured Reporting, sometimes called synoptic reporting, is a method of clinical documentation that captures and displays specific data elements within a particular format. In radiology, most reports are free-text narratives, which are variably formatted and prone to omitting essential data. In contrast, structured reporting templates provide consistency and clarity, prompt entry of all necessary data elements, and are amenable to scalable data capture, interoperability, and exchange.

Structured Reporting Templates: Most published structured reporting templates follow a technique that organizes data the body part examined. A variation, known as contextual reporting, tailors fields to the specific disease or examination indication, ensuring more relevant and precise documentation.

Most published examples of structured reporting templates are based on the technique and body part examined. A variant known as contextual reporting provides fields relevant to the disease or examination indication.

Teleradiology is a branch of telemedicine that revolves around transmitting radiological medical images in multiple modalities from one location to another to share studies with other radiologists and physicians for analysis and interpretation. As opposed to radiologists who work on-site, Teleradiologists interpret images and studies without being at the patient’s exact location. This flexibility is significant when a radiology sub-specialist such as an MRI radiologist, neuroradiologist, pediatric radiologist, or musculoskeletal radiologist, is needed, since these professionals are generally only located in large metropolitan areas working during daytime hours. Teleradiology allows trained specialists to be available 24/7.

Teleradiology workflows involve acquiring radiological images, transmitting them to remote radiologists for interpretation, and delivering diagnostic reports to the referring healthcare providers.

Turing Test: Originally called the imitation game by Alan Turing in 1949, the Turing Test is a test of whether a machine can think like a human. Turing proposed that a human evaluator would judge natural language conversations between humans and machines designed to generate human-like responses.

Unsupervised Learning: a form of machine learning where a machine intelligence processes unmoderated data samples and simply learns from whatever patterns and regularities it encounters.

Vendor Neutral Archive (VNA): a medical imaging technology in which images and documents (and potentially any file of clinical relevance) are stored (archived) in a standard format with a standard interface, such that they can be accessed in a vendor-neutral manner by other systems.

WADO-RS: is a DICOMweb term applied to the family of RESTful DICOM services defined for sending, retrieving, and querying medical images and related information. WADO-RS is for the retrieval of DICOM PS3.10 files, metadata in XML or JSON forms, bulk data separated from the metadata, and rendered consumer format images

Weak artificial intelligence (AI): a machine that mirrors human intelligence in a single field (such as deep analysis of data sets).