Imaging informatics, also known as radiology informatics or medical imaging informatics, is a subspecialty of biomedical informatics that aims to improve the efficiency, accuracy, usability and reliability of medical imaging services within the healthcare enterprise. [1] It is devoted to the study of how information about and contained within medical images is retrieved, analyzed, enhanced, and exchanged throughout the medical enterprise.
As radiology is an inherently data-intensive and technology-driven specialty, those in this branch of medicine have become leaders in Imaging Informatics. However, with the proliferation of digitized images across the practice of medicine to include fields such as cardiology, ophthalmology, dermatology, surgery, gastroenterology, obstetrics, gynecology and pathology, the advances in Imaging Informatics are also being tested and applied in other areas of medicine. Various industry players and vendors involved with medical imaging, along with IT experts and other biomedical informatics professionals, are contributing and getting involved in this expanding field.
Imaging informatics exists at the intersection of several broad fields:
Due to the diversity of the industry players and broad professional fields involved with Imaging Informatics, there grew a demand for new standards and protocols. These include DICOM (Digital Imaging and Communications in Medicine), Health Level 7 (HL7), International Organization for Standardization (ISO), and Artificial Intelligence protocols.
Current research surrounding Imaging Informatics has a focus on Artificial Intelligence (AI) and Machine Learning (ML). These new technologies are being used to develop automation methods, disease classification, advanced visualization techniques, and improvements in diagnostic accuracy. However, AI and ML integration faces several challenges with data management and security.
While the field of imaging informatics is based around the power of modern computing, its roots trace back to the dawn of the 20th century. On November 8th, 1895, German physicist Wilhelm Conrad Röntgen observed a new imaging technique he coined “X-rays” during his experiments. This discovery led to the creation of the medical imaging field, and in turn launched a new wave of human innovation. [2]
X-rays stood as the only medical imaging technology for several decades following its discovery. However, the arrival of the mid 20th century meant the expansion of the medical imaging field. The new modalities included: computed tomography (CT) to visualize soft tissue with a high degree of resolution; Magnetic Resonance Imaging (MRI) which is a modern standard for soft tissue imaging; Ultrasound that uses sound waves to create less expensive visualizations; Nuclear Imaging and Hybrid Scanners for functional imaging and imaging with higher spatial resolution created by combining multiple modalities. [3]
As these imaging techniques become more sophisticated, the amount of information medical imaging professionals are expected to process also increases. Additionally, the digital revolution of the mid to late 20th century further increased the data these techniques can gather. As a result, the main limiting factor for the medical imaging field became the human inability to accurately interpret large amounts of data. [4] Thus, the need arose for computerized assistance with complex digital imaging analysis, storage and manipulation. Modern Imaging Informatics was developed to fulfill these needs.
Imaging Informatics is a broad field with numerous areas of interest, making its development a culmination of the development of various individual technologies. Several of the key innovations for the field are as follows:
The development of PACS popularized the use of image storage and retrieval systems in medical practices. [5] Moreover, this new technology demanded the development of others. The world quickly realized that digital imaging standards would need to be put in place given the impact PACS had on the medical community. The American College of Radiology (ACR) and the National Electrical Manufacturers Association (NEMA) created the Digital Imaging and Communications Standards Committee (later becoming DICOM) in response to this concern. [6]
The digital age’s impact on radiology resulted in a large influx of data that needed to be managed. To combat this, the field of information technology was incorporated with technology such as Radiology Information System (RIS) [7] and Hospital Information System (HIS). These systems would work in tandem with PACS and other imaging technology to streamline the patient data management. [8]
The idea of computer-aided detection (CAD) and computer-aided diagnosis (CADx) is that the process of analysis and interpretation of medical image data could be automated, with a potentially higher degree of accuracy than human detection and diagnosis. Interest in this subject dates back to 1966, when radiology imaging first became digitized. [9] The first successful implementation of a CAD system was in 1994 at the University of Chicago for use in mammography. This was followed by the first commercial CAD system in 1998 called ImageChecker M1000. [6] With the arrival of the 21st century, machine learning techniques have been utilized to accomplish a version of the CAD and CADx systems. [10]
In the domain of imaging informatics, it is imperative to ascertain that the information pertaining to industry standards and data-sharing protocols is contemporaneous. The expeditious advancement in this field necessitates a vigilant approach to sustain uniformity, foster interoperability, and guarantee the efficacious dissemination of imaging data. To this end, several pivotal facets warrant rigorous consideration:
The Digital Imaging and Communications in Medicine (DICOM) standard delineates a sophisticated structural schema that integrates medical imaging data with pertinent patient identifiers into unified data sets, analogous to the embedded metadata in JPEG images. Such DICOM entities are constituted by a multitude of attributes, notably encapsulating pixel data, which in certain imaging modalities, corresponds to discrete images or, alternatively, an array of frames exemplifying kinetic or volumetric data, as observed in cine loops or multi-dimensional scans in nuclear medicine. This architecture accommodates the assimilation of intricate, multi-faceted data into a monolithic DICOM file. The standard accommodates a spectrum of pixel data compression algorithms, including but not limited to JPEG and JPEG 2000, and provisionally allows for holistic data set compression. DICOM specifies three encodings for data elements, with a predilection for explicit value representations, barring specific exceptions as elaborated in Part 5 of the DICOM compendium. Uniformly applied across diverse applications, the file manifestation customarily incorporates a header that houses essential attributes and data on the originating application.
The proposed workflow integrates the use of DICOM Structured Reporting (SR), in which essential measurements are encoded as DICOM SR objects. These objects are then utilized to fill a predefined SR template, resulting in the creation of a standardized report comprised of discrete data elements. This report is subsequently transmitted to the Electronic Medical Record (EMR) system. The discrete data extracted from these reports facilitate the longitudinal monitoring of individual patient metrics, are forwarded to data registries, or are leveraged for clinical research purposes. [11]
DDInteract has been crafted to enhance cooperative engagement between healthcare practitioners and patients, aiming to ascertain the optimal therapeutic approach that minimizes the hazards posed by potential drug-drug interactions. The user interface of DDInteract is systematically organized into four distinct segments.
Medication data can be represented across a variety of Fast Health Interoperability Resources (FHIR) resources, necessitating careful analysis by DDInteract. Specifically, MedicationRequest is utilized for medications prescribed to the patient; MedicationDispense covers medications that have been physically provided to the patient; and MedicationStatement pertains to medications that the patient reports having taken or is currently taking. It is possible for a single medication to be represented in multiple resource forms, with potential redundancies being amalgamated into a single record based on the most recent date and a defined hierarchy among the resource types.
To optimize the efficiency of data retrieval from the FHIR server, not every instance of medication is considered. Only those resources that are currently active or were active within the past 100 days are included, adhering to the prevalent U.S. protocol that typically allows for medication dispensation for a duration not exceeding three months.
A Quality Management System (QMS) is an integrative construct that includes the organizational architecture, the allocation of resources, the expertise of personnel, and the repository of documents and procedures that collectively contribute to the assurance and enhancement of quality in an entity's offerings. It delineates a suite of systematically orchestrated actions essential for governing and optimizing quality parameters. The ISO 9000 suite emerges as the preeminent and universally endorsed schema for QMS implementations, whereas the ISO 15189 standard provides a specialized framework designed expressly for the exigencies of clinical laboratory settings. [12]
A systematic review critically assessed the design, reporting standards, risk of bias, and validity of claims within studies that compare the efficacy of diagnostic deep learning algorithms in medical imaging against the expertise of clinicians. Conducted using data from prominent databases spanning from 2010 to June 2019, the review specifically targeted studies involving convolutional neural networks (CNNs)—notable for their capacity to autonomously discern crucial features for image classification within medical contexts. The investigation uncovered a notable deficiency in randomized clinical trials concerning this subject, identifying only ten such studies, of which merely two were published, exhibiting low risk of bias and commendable adherence to reporting protocols. Among the 81 non-randomized studies located, a minority were prospective or validated in practical clinical settings, with the majority presenting a high risk of bias, substandard compliance with reporting norms, and a pronounced lack of accessibility to data and code. This review underscores the imperative for an augmentation in the number of prospective studies and randomized trials, advocating for diminished bias, amplified clinical pertinence, enhanced transparency, and tempered conclusions in the burgeoning field of applying deep learning to medical imaging. [13]
The exponential growth in digital data alongside enhanced computing capabilities has markedly accelerated advancements in artificial intelligence (AI), which are now progressively being incorporated into healthcare. These AI applications aim to refine diagnosis, treatment, and prognosis through sophisticated classification and prediction models. Nevertheless, the evolution of these technologies is impeded by a lack of rigorous reporting standards relating to data sourcing, model architecture, and the methodologies employed in model evaluation and validation. In response, we propose MINIMAR (Minimum Information for Medical AI Reporting), an initiative designed to establish critical parameters for understanding AI-driven predictions, the demographics targeted, inherent biases, and the ability to generalize these technologies. We urge the adoption of standardized protocols to ensure that AI implementations in healthcare are reported with accuracy and responsibility, facilitating the development and deployment of associated clinical decision-support tools while simultaneously addressing critical concerns regarding precision and bias. [14]
As a foundational requisite, the proposed standard ought to fulfill several essential criteria: Firstly, it should encompass comprehensive details concerning the population from which the training data are derived, delineating the sources of data and the methods employed for cohort selection. Secondly, the demographics of the training data should be explicitly documented to facilitate a substantive comparison with the demographic characteristics of the population on which the model is intended to operate. Thirdly, there should be a thorough disclosure of the model’s architecture and its development process to allow for a clear interpretation of the model's intended purpose, comparison with analogous models, and to enable exact replication. Fourthly, the process of model evaluation, optimization, and validation must be transparently reported to elucidate the means by which local model optimization is attained and to support replication and the sharing of resources. [15]
Optimization | Model or parameter tuning applied | Generated vectors with a dimension of 300 and a window size of 5 | Documented and provided for all models in detail |
---|---|---|---|
Internal model validation | Study internal validation | Internal 10-fold cross-validation | Hold-out validation set |
External model validation | External validation using data from another setting | Not performed | Not performed |
Transparency | How code and data are shared with the community. | Code and sample data available via GitHub | Data is not available; code is available via GitHub |
In summary, while AI offers significant opportunities for advancing imaging informatics, leveraging these opportunities to their fullest extent necessitates stringent validation, adherence to robust reporting frameworks, and an overarching commitment to addressing ethical considerations. These steps are pivotal in ensuring that AI-driven tools achieve their promise of enhancing efficiency and effectiveness in medical diagnostics.
Key areas relevant to Imaging informatics include:
Imaging Informatics has quite a few applications within the medical field.
Imaging Informatics is most prominent within the field of radiology. Using AI, radiologists can use Imaging Informatics to ease their job and save time whilst analyzing images. A study published in "Current Medical Imaging" discovered that in CT imaging assisted by AI, the reading time to detect lung nodules and pleural effusions was reduced by more than 44% for radiologists. [18]
Imaging informatics within Cardiology aids in the molecular phenotyping of CV(Cardiovascular) diseases and unification of CV knowledge. [19] This means that through data extraction, imaging, and machine learning analysis of these data and images allow researchers to categorize diseases based on the characteristics or features discovered. With this classification, researchers are then able to unify this CV information into one platform for continued analysis and information retrieval.
Imaging informatics in pathology as a whole allows for a wide range of disease detection and analysis. The most prominent use in pathology is with the detection and analysis of different forms of cancer. Diagnosing cancer manually is a pain staking and subjective process which includes examining what could be millions of cells. Through various clinical decision support systems(CDSS), professionals can ease the manual labor of tissue region selection, using Whole-Slide Imaging(WSI) tools to maximize the information analyzed. Several predictive models aimed to identify regions of interest within WSI, requiring training before use. Unsupervised models are being introduced, but are currently less prominent. [20]
In the US and some other countries, radiologists who wish to pursue sub-specialty training in this field can undergo fellowship training in imaging informatics. Medical Imaging Informatics Fellowships are done after completion of Board Certification in Diagnostic Radiology, and may be pursued concurrently with other sub-specialty radiology fellowships.
The American Board of Imaging Informatics (ABII) also administers a certification examination for Imaging Informatics Professionals. PARCA (PACS Administrators Registry and Certification Association) certifications also exist for imaging informatics professionals. [21]
The American Board of Preventive Medicine (ABPM) offers a certification examination for Clinical Informatics for physicians who have primary board certification with the American Board of Medical Specialties, a medical license and a medical degree. There are two pathways to be eligible to sit for the examination: Practice Pathway (open through 2022) for those who have not completed ACGME-accredited fellowship training in Clinical Informatics and ACGME-Accredited Fellowship Pathway of at least 24 months in duration. [22]
Recent studies are focused on the application of artificial intelligence (AI) and machine learning (ML). [23] Areas of interest are disease classification and advanced visualization techniques. AI and ML are being implemented to improve diagnostic accuracy, automate processes, and develop personalized medicine. From an analysis of 562 highly cited papers. The "Chinese Academy of Sciences has the highest number of publications, while the institution with the highest centrality is Stanford University. The United States has the highest number of publications, followed by China. The most frequent keyword is “Deep Learning,” and the highest centrality keyword is “Algorithm.” The most cited author is Kaiming He, and the author with the highest centrality is Yoshua Bengio." [24] Deep Learning application is becoming common practice in the field of medical image processing. "Current research in medical image processing mainly focuses on deep learning, convolutional neural networks, classification, diagnosis, segmentation, image, algorithm, and artificial intelligence." [24] Trends show that research is becoming more complex and systematic. Deep Learning technology will play a critical role in the future.
There are several challenges in the field of Imaging Informatics:
Recent years have seen significant advancements in software technologies relevant to imaging informatics. One notable development is the integration of machine learning algorithms into imaging software, enabling automated analysis and interpretation of medical images. For instance, Rajpurkar et al. (2017) demonstrated the effectiveness of deep learning algorithms in pneumonia detection on chest X-rays, showcasing the potential of machine learning in medical imaging analysis. [27] These algorithms have shown promising results in tasks such as lesion detection, disease classification, and treatment response assessment. Moreover, the implementation of natural language processing (NLP) techniques has facilitated the extraction of valuable insights from unstructured radiology reports, enhancing the efficiency of data analysis and decision-making processes.
Advances in hardware technology have also played a pivotal role in shaping the landscape of imaging informatics. The evolution of imaging modalities, such as magnetic resonance imaging (MRI), computed tomography (CT), and positron emission tomography (PET), has led to improvements in image resolution, acquisition speed, and diagnostic accuracy [28] . Additionally, the miniaturization of imaging devices has enabled point-of-care imaging, allowing for real-time assessment of patients in various clinical settings. For example, the development of handheld ultrasound devices has revolutionized point-of-care imaging by providing clinicians with portable and easy-to-use tools for bedside examinations (Smith, 2018). The rise of wearable devices and mobile health applications has further expanded the scope of imaging informatics, facilitating remote imaging and patient monitoring using sensors and cameras [28] .
Along with technological innovations, methodological advancements have expanded the capabilities of imaging informatics. One development is the integration of multimodal imaging techniques, which combine data from multiple imaging modalities to provide complementary information about anatomical and physiological structures. For instance, recent studies have demonstrated the effectiveness of combining MRI, CT, and ultrasound data for improved diagnosis and treatment planning in oncology patients (Gupta et al., 2020) [29] . By fusing data from these sources, clinicians can obtain a more comprehensive understanding of a patient's condition, leading to more accurate diagnoses and personalized treatment plans.
A picture archiving and communication system (PACS) is a medical imaging technology which provides economical storage and convenient access to images from multiple modalities. 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, once encapsulated in DICOM. A PACS consists of four major components: The imaging modalities such as X-ray plain film (PF), computed tomography (CT) and magnetic resonance imaging (MRI), a secured network for the transmission of patient information, workstations for interpreting and reviewing images, and archives for the storage and retrieval of images and reports. Combined with available and emerging web technology, PACS has the ability to deliver timely and efficient access to images, interpretations, and related data. PACS reduces the physical and time barriers associated with traditional film-based image retrieval, distribution, and display.
Digital Imaging and Communications in Medicine (DICOM) is a technical standard for the digital storage and transmission of medical images and related information. It includes a file format definition, which specifies the structure of a DICOM file, as well as a network communication protocol that uses TCP/IP to communicate between systems. The primary purpose of the standard is to facilitate communication between the software and hardware entities involved in medical imaging, especially those that are created by different manufacturers. Entities that utilize DICOM files include components of picture archiving and communication systems (PACS), such as imaging machines (modalities), radiological information systems (RIS), scanners, printers, computing servers, and networking hardware.
Radiology is the medical specialty that uses medical imaging to diagnose diseases and guide their treatment, within the bodies of humans and other animals. It began with radiography, but today it includes all imaging modalities, including those that use no ionizing electromagnetic radiation, as well as others that do, such as computed tomography (CT), fluoroscopy, and nuclear medicine including positron emission tomography (PET). Interventional radiology is the performance of usually minimally invasive medical procedures with the guidance of imaging technologies such as those mentioned above.
Medical physics deals with the application of the concepts and methods of physics to the prevention, diagnosis and treatment of human diseases with a specific goal of improving human health and well-being. Since 2008, medical physics has been included as a health profession according to International Standard Classification of Occupation of the International Labour Organization.
Machine learning (ML) is a field of study in artificial intelligence concerned with the development and study of statistical algorithms that can learn from data and generalize to unseen data, and thus perform tasks without explicit instructions. Recently, artificial neural networks have been able to surpass many previous approaches in performance.
Medical imaging is the technique and process of imaging the interior of a body for clinical analysis and medical intervention, as well as visual representation of the function of some organs or tissues (physiology). Medical imaging seeks to reveal internal structures hidden by the skin and bones, as well as to diagnose and treat disease. Medical imaging also establishes a database of normal anatomy and physiology to make it possible to identify abnormalities. Although imaging of removed organs and tissues can be performed for medical reasons, such procedures are usually considered part of pathology instead of medical imaging.
Health informatics is the study and implementation of computer structures and algorithms to improve communication, understanding, and management of medical information. It can be viewed as branch of engineering and applied science.
An electronic health record (EHR) is the systematized collection of patient and population electronically stored health information in a digital format. These records can be shared across different health care settings. Records are shared through network-connected, enterprise-wide information systems or other information networks and exchanges. EHRs may include a range of data, including demographics, medical history, medication and allergies, immunization status, laboratory test results, radiology images, vital signs, personal statistics like age and weight, and billing information.
A clinical decision support system (CDSS) is a health information technology that provides clinicians, staff, patients, and other individuals with knowledge and person-specific information to help health and health care. CDSS encompasses a variety of tools to enhance decision-making in the clinical workflow. These tools include computerized alerts and reminders to care providers and patients, clinical guidelines, condition-specific order sets, focused patient data reports and summaries, documentation templates, diagnostic support, and contextually relevant reference information, among other tools. CDSSs constitute a major topic in artificial intelligence in medicine.
Personalized medicine, also referred to as precision medicine, is a medical model that separates people into different groups—with medical decisions, practices, interventions and/or products being tailored to the individual patient based on their predicted response or risk of disease. The terms personalized medicine, precision medicine, stratified medicine and P4 medicine are used interchangeably to describe this concept, though some authors and organizations differentiate between these expressions based on particular nuances. P4 is short for "predictive, preventive, personalised and participatory".
Computer-aided detection (CADe), also called computer-aided diagnosis (CADx), are systems that assist doctors in the interpretation of medical images. Imaging techniques in X-ray, MRI, Endoscopy, and ultrasound diagnostics yield a great deal of information that the radiologist or other medical professional has to analyze and evaluate comprehensively in a short time. CAD systems process digital images or videos for typical appearances and to highlight conspicuous sections, such as possible diseases, in order to offer input to support a decision taken by the professional.
Digital pathology is a sub-field of pathology that focuses on data management based on information generated from digitized specimen slides. Through the use of computer-based technology, digital pathology utilizes virtual microscopy. Glass slides are converted into digital slides that can be viewed, managed, shared and analyzed on a computer monitor. With the practice of whole-slide imaging (WSI), which is another name for virtual microscopy, the field of digital pathology is growing and has applications in diagnostic medicine, with the goal of achieving efficient and cheaper diagnoses, prognosis, and prediction of diseases due to the success in machine learning and artificial intelligence in healthcare.
Health information technology (HIT) is health technology, particularly information technology, applied to health and health care. It supports health information management across computerized systems and the secure exchange of health information between consumers, providers, payers, and quality monitors. Based on a 2008 report on a small series of studies conducted at four sites that provide ambulatory care – three U.S. medical centers and one in the Netherlands, the use of electronic health records (EHRs) was viewed as the most promising tool for improving the overall quality, safety and efficiency of the health delivery system.
Computers and software have been used in dental medicine since the 1960s. Since then, computers and information technology have spread progressively in dental practice. According to one study, in 2000, 85.1% of all dentists in the United States were using computers.
Medical image sharing is the electronic exchange of medical images between hospitals, physicians and patients. Rather than using traditional media, such as a CD or DVD, and either shipping it out or having patients carry it with them, technology now allows for the sharing of these images using the cloud. The primary format for images is DICOM. Typically, non-image data such as reports may be attached in standard formats like PDF during the sending process. Additionally, there are standards in the industry, such as IHE Cross Enterprise Document Sharing for Imaging (XDS-I), for managing the sharing of documents between healthcare enterprises. A typical architecture involved in setup is a locally installed server, which sits behind the firewall, allowing secure transmissions with outside facilities. In 2009, the Radiological Society of North America launched the "Image Share" project, with the goal of giving patients control of their imaging histories by allowing them to manage these records as they would online banking or shopping.
Artificial intelligence in healthcare is a term used to describe the use of machine-learning algorithms and software, or artificial intelligence (AI), to copy human cognition in the analysis, presentation, and understanding of complex medical and health care data, or to exceed human capabilities by providing new ways to diagnose, treat, or prevent disease. Specifically, AI is the ability of computer algorithms to arrive at approximate conclusions based solely on input data.
Georgia "Gina" D. Tourassi is the Director of the Oak Ridge National Laboratory health data sciences institute and adjunct Professor of radiology at Duke University. She works on biomedical informatics, computer-aided diagnosis and artificial intelligence (AI) in health care.
Aidoc Medical is an Israeli technology company that develops computer-aided simple triage and notification systems. Aidoc has obtained FDA and CE mark approval for its stroke, pulmonary embolism, cervical fracture, intracranial hemorrhage, intra-abdominal free gas, and incidental pulmonary embolism algorithms.
Merative L.P., formerly IBM Watson Health, is an American medical technology company that provides products and services that help clients facilitate medical research, clinical research, real world evidence, and healthcare services, through the use of artificial intelligence, data analytics, cloud computing, and other advanced information technology. Merative is owned by Francisco Partners, an American private equity firm headquartered in San Francisco, California. In 2022, IBM divested and spun-off their Watson Health division into Merative. As of 2023, it remains a standalone company headquartered in Ann Arbor with innovation centers in Hyderabad, Bengaluru, and Chennai.
Artificial intelligence (AI) in mental health refers to the use of AI, computational technologies and algorithms to supplement the understanding, diagnosis, and treatment of mental health disorders.