Ethics of Artificial Intelligence in Radiology: Summary of the Joint European and North American Multisociety Statement.

This is a condensed summary of an international multisociety statement on ethics of artificial intelligence (AI) in radiology produced by the ACR, European Society of Radiology, RSNA, Society for Imaging Informatics in Medicine, European Society of Medical Imaging Informatics, Canadian Association of Radiologists, and American Association of Physicists in Medicine. AI has great potential to increase efficiency and accuracy throughout radiology, but it also carries inherent pitfalls and biases. Widespread use of AI-based intelligent and autonomous systems in radiology can increase the risk of systemic errors with high consequence and highlights complex ethical and societal issues. Currently, there is little experience using AI for patient care in diverse clinical settings. Extensive research is needed to understand how to best deploy AI in clinical practice. This statement highlights our consensus that ethical use of AI in radiology should promote well-being, minimize harm, and ensure that the benefits and harms are distributed among stakeholders in a just manner. We believe AI should respect human rights and freedoms, including dignity and privacy. It should be designed for maximum transparency and dependability. Ultimate responsibility and accountability for AI remains with its human designers and operators for the foreseeable future. The radiology community should start now to develop codes of ethics and practice for AI that promote any use that helps patients and the common good and should block use of radiology data and algorithms for financial gain without those two attributes. This article is a simultaneous joint publication in Radiology, Journal of the American College of Radiology, Canadian Association of Radiologists Journal, and Insights into Imaging. Published under a CC BY-NC-ND 4.0 license. Online supplemental material is available for this article.

Correction to: Collaborative and Reproducible Research: Goals, Challenges, and Strategies.

The paper below had been published originally without open access, but has been republished with open access.

Ethics of Artificial Intelligence in Radiology: Summary of the Joint European and North American Multisociety Statement.

This is a condensed summary of an international multisociety statement on ethics of artificial intelligence (AI) in radiology produced by the ACR, European Society of Radiology, RSNA, Society for Imaging Informatics in Medicine, European Society of Medical Imaging Informatics, Canadian Association of Radiologists, and American Association of Physicists in Medicine. AI has great potential to increase efficiency and accuracy throughout radiology, but it also carries inherent pitfalls and biases. Widespread use of AI-based intelligent and autonomous systems in radiology can increase the risk of systemic errors with high consequence and highlights complex ethical and societal issues. Currently, there is little experience using AI for patient care in diverse clinical settings. Extensive research is needed to understand how to best deploy AI in clinical practice. This statement highlights our consensus that ethical use of AI in radiology should promote well-being, minimize harm, and ensure that the benefits and harms are distributed among stakeholders in a just manner. We believe AI should respect human rights and freedoms, including dignity and privacy. It should be designed for maximum transparency and dependability. Ultimate responsibility and accountability for AI remains with its human designers and operators for the foreseeable future. The radiology community should start now to develop codes of ethics and practice for AI that promote any use that helps patients and the common good and should block use of radiology data and algorithms for financial gain without those two attributes.

Collaborative and Reproducible Research: Goals, Challenges, and Strategies.

Combining imaging biomarkers with genomic and clinical phenotype data is the foundation of precision medicine research efforts. Yet, biomedical imaging research requires unique infrastructure compared with principally text-driven clinical electronic medical record (EMR) data. The issues are related to the binary nature of the file format and transport mechanism for medical images as well as the post-processing image segmentation and registration needed to combine anatomical and physiological imaging data sources. The SiiM Machine Learning Committee was formed to analyze the gaps and challenges surrounding research into machine learning in medical imaging and to find ways to mitigate these issues. At the 2017 annual meeting, a whiteboard session was held to rank the most pressing issues and develop strategies to meet them. The results, and further reflections, are summarized in this paper.

Learning HL7 FHIR Using the HAPI FHIR Server and Its Use in Medical Imaging with the SIIM Dataset.

Health Level 7's (HL7's) new standard, FHIR (Fast Health Interoperability Resources), is setting healthcare information technology and medical imaging specifically ablaze with excitement. This paper aims to describe the protocol's advantages in some detail and explore an easy path for those unfamiliar with FHIR to begin learning the standard using free, open-source tools, namely the HL7 application programming interface (HAPI) FHIR server and the SIIM Hackathon Dataset.

Creation and Curation of the Society of Imaging Informatics in Medicine Hackathon Dataset.

In order to support innovation, the Society of Imaging Informatics in Medicine (SIIM) elected to create a collaborative computing experience called a "hackathon." The SIIM Hackathon has always consisted of two components, the event itself and the infrastructure and resources provided to the participants. In 2014, SIIM provided a collection of servers to participants during the annual meeting. After initial server setup, it was clear that clinical and imaging "test" data were also needed in order to create useful applications. We outline the goals, thought process, and execution behind the creation and maintenance of the clinical and imaging data used to create DICOM and FHIR Hackathon resources.

Cyber-Security Issues in Healthcare Information Technology.

In 1999-2003, SIIM (then SCAR) sponsored the creation of several special topic Primers, one of which was concerned with computer security. About the same time, a multi-society collaboration authored an ACR Guideline with a similar plot; the latter has recently been updated. The motivation for these efforts was the launch of Health Information Portability and Accountability Act (HIPAA). That legislation directed care providers to enable the portability of patient medical records across authorized medical centers, while simultaneously protecting patient confidentiality among unauthorized agents. These policy requirements resulted in the creation of numerous technical solutions which the above documents described. While the mathematical concepts and algorithms in those papers are as valid today as they were then, recent increases in the complexity of computer criminal applications (and defensive countermeasures) and the pervasiveness of Internet connected devices have raised the bar. This work examines how a medical center can adapt to these evolving threats.

Workflow Lexicons in Healthcare: Validation of the SWIM Lexicon.

For clinical departments seeking to successfully navigate the challenges of modern health reform, obtaining access to operational and clinical data to establish and sustain goals for improving quality is essential. More broadly, health delivery organizations are also seeking to understand performance across multiple facilities and often across multiple electronic medical record (EMR) systems. Interpreting operational data across multiple vendor systems can be challenging, as various manufacturers may describe different departmental workflow steps in different ways and sometimes even within a single vendor's installed customer base. In 2012, The Society for Imaging Informatics in Medicine (SIIM) recognized the need for better quality and performance data standards and formed SIIM's Workflow Initiative for Medicine (SWIM), an initiative designed to consistently describe workflow steps in radiology departments as well as defining operational quality metrics. The SWIM lexicon was published as a working model to describe operational workflow steps and quality measures. We measured the prevalence of the SWIM lexicon workflow steps in both academic and community radiology environments using real-world patient observations and correlated that information with automatically captured workflow steps from our clinical information systems. Our goal was to measure frequency of occurrence of workflow steps identified by the SWIM lexicon in a real-world clinical setting, as well as to correlate how accurately departmental information systems captured patient flow through our health facility.

DICOM Data Warehouse: Part 2.

In 2010, the DICOM Data Warehouse (DDW) was launched as a data warehouse for DICOM meta-data. Its chief design goals were to have a flexible database schema that enabled it to index standard patient and study information, modality specific tags (public and private), and create a framework to derive computable information (derived tags) from the former items. Furthermore, it was to map the above information to an internally standard lexicon that enables a non-DICOM savvy programmer to write standard SQL queries and retrieve the equivalent data from a cohort of scanners, regardless of what tag that data element was found in over the changing epochs of DICOM and ensuing migration of elements from private to public tags. After 5 years, the original design has scaled astonishingly well. Very little has changed in the database schema. The knowledge base is now fluent in over 90 device types. Also, additional stored procedures have been written to compute data that is derivable from standard or mapped tags. Finally, an early concern is that the system would not be able to address the variability DICOM-SR objects has been addressed. As of this writing the system is indexing 300 MR, 600 CT, and 2000 other (XA, DR, CR, MG) imaging studies per day. The only remaining issue to be solved is the case for tags that were not prospectively indexed-and indeed, this final challenge may lead to a noSQL, big data, approach in a subsequent version.

Sensitivity of Thoracic Digital Tomosynthesis (DTS) for the Identification of Lung Nodules.

Thoracic computed tomography (CT) is considered the gold standard for detection lung pathology, yet its efficacy as a screening tool in regards to cost and radiation dose continues to evolve. Chest radiography (CXR) remains a useful and ubiquitous tool for detection and characterization of pulmonary pathology, but reduced sensitivity and specificity compared to CT. This prospective, blinded study compares the sensitivity of digital tomosynthesis (DTS), to that of CT and CXR for the identification and characterization of lung nodules. Ninety-five outpatients received a posteroanterior (PA) and lateral CXR, DTS, and chest CT at one care episode. The CXR and DTS studies were independently interpreted by three thoracic radiologists. The CT studies were used as the gold standard and read by a fourth thoracic radiologist. Nodules were characterized by presence, location, size, and composition. The agreement between observers and the effective radiation dose for each modality was objectively calculated. One hundred forty-five nodules of greatest diameter larger than 4 mm and 215 nodules less than 4 mm were identified by CT. DTS identified significantly more >4 mm nodules than CXR (DTS 32 % vs. CXR 17 %). CXR and DTS showed no significant difference in the ability to identify the smaller nodules or central nodules within 3 cm of the hilum. DTS outperformed CXR in identifying pleural nodules and those nodules located greater than 3 cm from the hilum. Average radiation dose for CXR, DTS, and CT were 0.10, 0.21, and 6.8 mSv, respectively. Thoracic digital tomosynthesis requires significantly less radiation dose than CT and nearly doubles the sensitivity of that of CXR for the identification of lung nodules greater than 4 mm. However, sensitivity and specificity for detection and characterization of lung nodules remains substantially less than CT. The apparent benefits over CXR, low cost, rapid acquisition, and minimal radiation dose of thoracic DTS suggest that it may be a useful procedure. Work-up of a newly diagnosed nodule will likely require CT, given its superior cross-sectional characterization. Further investigation of DTS as a diagnostic, screening, and surveillance tool is warranted.

Implementing a radiology-information technology project: mobile image viewing use case and a general guideline for radiologist-information technology team collaboration.

OBJECTIVE: This article illustrates the importance of radiologist engagement in the successful implementation of radiology-information technology (IT) projects through the example of establishing a mobile image viewing solution for health care professionals. CONCLUSION: With an understanding of the types of decisions that benefit from radiologist input, this article outlines an overall project framework to provide a context for how radiologists might engage in the project cycle.

MIRMAID: A Content Management System for Medical Image Analysis Research.

Today, a typical clinical study can involve thousands of participants, with imaging data acquired over several time points across multiple institutions. The additional associated information (metadata) accompanying these data can cause data management to be a study-hindering bottleneck. Consistent data management is crucial for large-scale modern clinical imaging research studies. If the study is to be used for regulatory submissions, such systems must be able to meet regulatory compliance requirements for systems that manage clinical image trials, including protecting patient privacy. Our aim was to develop a system to address these needs by leveraging the capabilities of an open-source content management system (CMS) that has a highly configurable workflow; has a single interface that can store, manage, and retrieve imaging-based studies; and can handle the requirement for data auditing and project management. We developed a Web-accessible CMS for medical images called Medical Imaging Research Management and Associated Information Database (MIRMAID). From its inception, MIRMAID was developed to be highly flexible and to meet the needs of diverse studies. It fulfills the need for a complete system for medical imaging research management.

DEWEY: the DICOM-enabled workflow engine system.

Workflow is a widely used term to describe the sequence of steps to accomplish a task. The use of workflow technology in medicine and medical imaging in particular is limited. In this article, we describe the application of a workflow engine to improve workflow in a radiology department. We implemented a DICOM-enabled workflow engine system in our department. We designed it in a way to allow for scalability, reliability, and flexibility. We implemented several workflows, including one that replaced an existing manual workflow and measured the number of examinations prepared in time without and with the workflow system. The system significantly increased the number of examinations prepared in time for clinical review compared to human effort. It also met the design goals defined at its outset. Workflow engines appear to have value as ways to efficiently assure that complex workflows are completed in a timely fashion.

Visualizing Clinical Data Retrieval and Curation in Multimodal Healthcare AI Research: A Technical Note on RIL-workflow.

Curating and integrating data from sources are bottlenecks to procuring robust training datasets for artificial intelligence (AI) models in healthcare. While numerous applications can process discrete types of clinical data, it is still time-consuming to integrate heterogenous data types. Therefore, there exists a need for more efficient retrieval and storage of curated patient data from dissimilar sources, such as biobanks, health records, and sensors. We describe a customizable, modular data retrieval application (RIL-workflow), which integrates clinical notes, images, and prescription data, and show its feasibility applied to research at our institution. It uses the workflow automation platform Camunda (Camunda Services GmbH, Berlin, Germany) to collect internal data from Fast Healthcare Interoperability Resources (FHIR) and Digital Imaging and Communications in Medicine (DICOM) sources. Using the web-based graphical user interface (GUI), the workflow runs tasks to completion according to visual representation, retrieving and storing results for patients meeting study inclusion criteria while segregating errors for human review. We showcase RIL-workflow with its library of ready-to-use modules, enabling researchers to specify human input or automation at fixed steps. We validated our workflow by demonstrating its capability to aggregate, curate, and handle errors related to data from multiple sources to generate a multimodal database for clinical AI research. Further, we solicited user feedback to highlight the pros and cons associated with RIL-workflow. The source code is available at github.com/magnooj/RIL-workflow.

Validation and initial clinical use of automatic peak skin dose localization with fluoroscopic and interventional procedures.

PURPOSE: To assess the accuracy and initial clinical use of a software tool that automatically maps and records values of skin dose, including peak skin dose (PSD), administered to patients undergoing fluoroscopically guided interventional procedures. MATERIALS AND METHODS: In this retrospective study, the institutional review board determined that this HIPAA-compliant study met the criteria as a quality assurance investigation. Informed consent was waived. After the initial validation and accuracy tests, distributed skin dose and PSD estimates were obtained for fluoroscopically guided interventional procedures performed in the radiology, cardiology, and gastroenterology practice areas between January and October 2011. A total of 605 procedures were performed in 520 patients (64% men; age range, 20-95 years). The accuracy of a skin dose tool to estimate patient dose distribution was verified with phantom studies by using an external dosimeter and direct exposure film. PSD distribution, PSD according to procedure type, and PSD for individual physician operators were assessed. RESULTS: Calculated PSD values agreed within ±9% of that measured by using film dosimetry under the condition of matched-phantom geometry. The area receiving the highest dose (greater than 95% of peak) agreed within ±17%. Of 605 patient procedures, 15 demonstrated PSD greater than 2 Gy, with a maximum PSD of 5.6 Gy. CONCLUSION: Knowledge of the patient skin dose can help direct treatment of patients who were administered relatively high skin dose and may be used to plan future procedures. SUPPLEMENTAL MATERIAL: http://radiology.rsna.org/lookup/suppl/doi:10.1148/radiol.12112295/-/DC1.

Medical imaging displays and their use in image interpretation.

The adequate and repeatable performance of the image display system is a key element of information technology platforms in a modern radiology department. However, despite the wide availability of high-end computing platforms and advanced color and gray-scale monitors, the quality and properties of the final displayed medical image may often be inadequate for diagnostic purposes if the displays are not configured and maintained properly. In this article-an expanded version of the Radiological Society of North America educational module "Image Display"-the authors discuss fundamentals of image display hardware, quality control and quality assurance processes for optimal image interpretation settings, and parameters of the viewing environment that influence reader performance. Radiologists, medical physicists, and other allied professionals should strive to understand the role of display technology and proper usage for a quality radiology practice. The display settings and display quality control and quality assurance processes described in this article can help ensure high standards of perceived image quality and image interpretation accuracy.

Standards for business analytics and departmental workflow.

Efficient workflow is essential for a successful business. However, there is relatively little literature on analytical tools and standards for defining workflow and measuring workflow efficiency. Here, we describe an effort to define a workflow lexicon for medical imaging departments, including the rationale, the process, and the resulting lexicon.

Towards a more cloud-friendly medical imaging applications architecture: a modest proposal.

Recent information technology literature, in general, and radiology trade journals, in particular, are rife with allusions to the "cloud" suggesting that moving one's compute and storage assets into someone else's data center magically solves cost, performance, and elasticity problems. More likely, one is only trading one set of problems for another, including greater latency (aka slower turnaround times) since the image data must now leave the local area network and travel longer paths via encrypted tunnels. To offset this, an imaging system design is needed that reduces the number of high-latency image transmissions, yet can still leverage cloud strengths. This work explores the requirements for such a design.

Human contrast-detail performance with declining contrast.

PURPOSE: How do display settings and ambient lighting affect contrast detection thresholds for human observers? Can recalibrating a display for high ambient lighting improve object detection? METHODS: Contrast∕detail (CD) threshold detection performance was measured for observers using four color displays with varying overall contrast (e.g., differing maximum luminance and ambient lighting conditions). Detailed mapping of contrast detection performance (for fixed object size) was tracked as a function of: display maximum luminance, ambient lighting changes (with and without recalibrating for the higher ambience), and the performance of radiologists vs. nonradiologists. RESULTS: The initial phase was analyzed with a hierarchical linear model of observer performance using: background gray level, maximum display luminance, and radiologist vs. nonradiologist. The only statistically significant finding was a maximum luminance of 100 cd∕m(2) display performing worse than a baseline peak of 400 cd∕m(2). The second phase examined ambient lighting effects on detection thresholds. Background gray level and maximum display luminance were examined coupled with ambient lighting for: baseline at 30, 435 uncorrected, and 435 lx with display recalibration for the ambient conditions. Results showed ambient correction improved sensitivity for small background digital driving level, but not at higher luminance backgrounds. CONCLUSIONS: For CD study, nonradiologist observers can be used without loss of applicability. Contrast detection thresholds improved significantly between displays with peak luminance from 100 cd∕m(2) to 200 cd∕m(2), but improvement beyond that was not statistically significant for contrast detection thresholds in a reading room environment. Applying a calibration correction at high ambience (435 lx) improved detection tasks primarily in the darker background regions.