REVOLUTIONIZING RADIOLOGY: A SYSTEMATIC REVIEW OF HOW ARTIFICIAL INTELLIGENCE CAN IMPACT THE AMERICAN HEALTHCARE SYSTEM By Nicki Dangleis Kristin Duncan Benjamin Ewing Jama Hassan Logan Kurtek Kelli McBee Kyndal Mick Jay Moffett Keri Phillips Etilia Riquelme Holli Schock Lynn Smith Cassandra Verzwyvelt A thesis submitted to the School of Radiologic Sciences in collaboration with a research agenda team In partial fulfillment of the requirements for the degree of MASTER OF SCIENCE IN RADIOLOGIC SCIENCES (MSRS) WEBER STATE UNIVERSITY Ogden, Utah December 13, 2024 THE WEBER STATE UNIVERSITY GRADUATE SCHOOL SUPERVISORY COMMITTEE APPROVAL of a thesis submitted by Nicki Dangleis Kristin Duncan Benjamin Ewing Jama Hassan Logan Kurtek Kelli McBee Kyndal Mick Jay Moffett Keri Phillips Etilia Riquelme Holli Schock Lynn Smith Cassandra Verzwyvelt This thesis has been read by each member of the following supervisory committee and by majority vote found to be satisfactory. ______________________________ Dr. Tanya Nolan, EdD Chair, School of Radiologic Sciences ______________________________ Dr. Robert Walker, PhD Director of MSRS ______________________________ Dr. Laurie Coburn, EdD Director of MSRS RA ______________________________ Christopher Steelman, MS Director of MSRS Cardiac Specialist THE WEBER STATE UNIVERSITY GRADUATE SCHOOL RESEARCH AGENDA STUDENT APPROVAL of a thesis submitted by Nicki Dangleis Kristin Duncan Benjamin Ewing Jama Hassan Logan Kurtek Kelli McBee Kyndal Mick Jay Moffett Keri Phillips Etilia Riquelme Holli Schock Lynn Smith Cassandra Verzwyvelt This thesis has been read by each member of the student research agenda committee and by majority vote found to be satisfactory. Date December 13, 2024 _____________________ December 13, 2024 ______________________ December 13, 2024 ______________________ ____________________________________ Nicki Dangleis ____________________________________ Kristin Duncan ____________________________________ Benjamin Ewing December 13, 2024 ______________________ December 13, 2024 ______________________ December 13, 2024 ______________________ December 13, 2024 ______________________ December 13, 2024 ______________________ December 13, 2024 ______________________ December 13, 2024 ______________________ December 13, 2024 ______________________ December 13, 2024 ______________________ December 13, 2024 ______________________ ____________________________________ Jama Hassan ____________________________________ Logan Kurtek ____________________________________ Kelli McBee ____________________________________ Kyndal Mick ____________________________________ Jay Moffett ____________________________________ Keri Phillips ____________________________________ Etilia Riquelme ____________________________________ Holli Schock ____________________________________ Lynn Smith ____________________________________ Cassandra Verzwyvelt Abstract The use of artificial intelligence (AI) in radiology marks a significant breakthrough for the future of medical imaging technology. As AI technology advances and becomes more integrated into healthcare, its influence on radiology continues to evolve at a rapid pace. However, despite the rapid growth, the United States seems to be lagging in AI implementation in comparison to other medically advanced countries. This systematic review provides a qualitative analysis of the current AI landscape in imaging, covering present uses of AI technology, as well as its successes and limitations. Articles and studies from both the United States and around the world reveal multiple countries are effectively improving patient care. Quantitative data revealed a 60% reduction in missed diagnoses of lung conditions through deep learning algorithms analyzing chest radiographs. Additionally, AI increased processing speed by 21%, notably improving urgent care scenarios like stroke assessments, thereby expediting the decision-making process for critical interventions. These results show that while AI is revolutionizing radiology by enhancing diagnostic speed and accuracy, its integration also demands careful handling of challenges such as ethical concerns, legal framework, financial hesitations, and shifts in departmental roles. Looking ahead, it is essential to focus on developing educational programs, enhancing data-sharing practices, and setting ethical standards to fully leverage AI's potential in transforming radiology and improving patient care. Future research should not only validate and improve AI's capabilities but also create strong frameworks to allow proper implementation and growth. Acknowledgements This thesis would not have been possible without the guidance, support, and encouragement of many people, to whom we are deeply grateful. First and foremost, we extend our sincere thanks to our director, Dr. Laurie Coburn, for her invaluable guidance, mentorship, and patience throughout this research process. Dr. Coburn’s expertise and insight has greatly enriched this work and our academic growth. Our deepest gratitude goes to our family for their unconditional love, understanding, and encouragement throughout this journey. Their support has been our foundation and source of strength, allowing us to persevere through challenges and celebrate accomplishments. We also wish to acknowledge our colleagues and friends who have offered support, feedback, and camaraderie. Their encouragement and insightful discussions have been invaluable in refining our ideas and approaches. Finally, we would like to express our appreciation to the faculty and staff of Weber State University, whose commitment to academic excellence and student success has made this journey both possible and fulfilling. Thank you all for your contributions to this endeavor. This thesis is a testament to your encouragement and guidance. Table of Contents Chapter 1: Introduction…………………………………………………………………………... 1 Background………………………………………………………………………………. 1 Statement of the Problem………………………………………………………………… 2 Purpose of the Study……………………………………………………………………... 3 Methods …………...……………………………………………………………………... 3 Research Questions………………………………………………………………………. 4 Nature of the Study………………………………………………………………………. 7 Significance of the Study………………………………………………………………… 7 Definition of Key Terms…………………………………………………………………. 8 Summary…………………………………………………………………………………. 9 Chapter 2: Literature Review…………………………………………………………………….11 Documentation……………………………………………………………………….…..11 Introduction and Background……………………………………………………………11 Accuracy and Efficiency…………………………………………………………………13 Uses Overseas……………………………………………………………………………16 Uses in the USA…………………………………………………………………….……19 Legalities/Ethics…………………………………………………….……………………24 Future of AI………………………………………………………………………...…….26 Summary…………………………………………………………………………………28 Chapter 3: Research Method……………………………………………………………………..30 Research Methods and Design(s).......................................................................................34 Population………………………………………………………………………………..34 Sample………………………………………………………………………………...….35 Data Collection, Processing, and Analysis……………………………………………....35 Assumptions…………………………………………………………………………….. 36 Limitations……………………………………………………………………………….36 Delimitations……………………………………………………………………………..37 Ethical Assurances……………………………………………………………………….38 Summary…………………………………………………………………………………38 Chapter 4: Findings………………………………………………………………………………40 Results……………………………………………………………………………………40 Evaluation of Findings…………………………………………………………………...41 Summary…………………………………………………………………………………42 Chapter 5: Implications, Recommendations, and Conclusions………………………………….44 Implications………………………………………………………………………………44 Recommendations………………………………………………………………………..45 Conclusions………………………………………………………………………………46 References………………………………………………………………………………………..48 Appendices……………………………………………………………………………………….54 Appendix A: Screening Flow Chart per 2020 PRISMA Guidelines…………………….54 Appendix B: Highlighted Mammographic Abnormalities Detected by MIA…………...55 Appendix C: AI Integration of Red Dot®CXR in NHS Lung Cancer Pathway………...56 Appendix D: Red Dot®CXR Accelerating Diagnostics with Heatmap Insights………..57 Appendix E: Lung Analysis with Siemens' CT Pneumonia Analysis2………………….58 Appendix F: BMD Reporting with LiberaBMD Using Hologic DXA Data…………….59 Appendix G: GE CardIQ Suite Utilizing AI for CAD Detection………………………..60 Appendix H: Viz.ai LVO Detection and Workflow Optimization………………………61 List of Tables Table 1. Viz.ai Software analysis in assessing Large Vessel and Internal Carotid Artery or middle cerebral artery first segment occlusions………………………………………………………… 23 1 Chapter 1: Introduction For many years, advancements in radiology and computer technology have tightly intertwined. However, the compounding growth of Artificial Intelligence (AI) applications is quickly changing this relationship. According to a recent article published in Diagnostic and Interventional Imaging, researchers estimated that radiologists must review one image every 3-4 seconds to keep pace with current workload demands (Barat et al., 2023). With the help of groundbreaking AI applications, there is hope for managing such daunting workloads. Nearly every medical journal, magazine, and biomedical manufacturer is touting the latest and greatest AI innovation. Be that as it may, with so much information available, staying up-to-date with the constant developments in AI can be equally daunting. For this reason, this study provides a deepdive review into the current uses, effectiveness, legal and ethical considerations, and future trajectory of AI in Radiology. Background Many industries are actively implementing AI applications as the world increasingly embraces data-driven approaches. In just the past few years, the medical AI market has exploded, likely forever changing healthcare, with radiology at the forefront. For example, across the United Kingdom, approximately 100 AI applications are currently in use, such as the Behold.ai red dot® prioritization platform through Dartford and Gravesham NHS Trust, MIA in the Midlands, and OPTIMAM by DeepMind AI at Royal Surrey County Hospital (Mudgal et al., 2020). AI systems can increase workflow efficiency for both technologists and radiologists alike. Through analysis of the intricate imaging patterns, AI has the potential to enhance diagnostic accuracy, which can increase patient outcomes. 2 Statement of the Problem In contrast to similar nations, integrating AI into radiology departments in the United States is a crucial area of inquiry in contemporary healthcare. By focusing on AI's potential to enhance diagnostic accuracy, workflow efficiency, and patient care outcomes, our research aims to provide insights into the transformative capabilities of AI. We will compare the ethical boundaries of AI implementation, discuss compliance with regulations such as HIPAA, and address concerns regarding patient privacy and data security. Our research aims to shed light on the challenges confronting radiologists in handling augmented workloads while examining how AI can optimize workflows, enhance productivity, and mitigate expenses linked to unnecessary medical interventions. By highlighting the economic benefits of AI integration, particularly the expected decrease in healthcare expenditures in the United States, our findings will emphasize the necessity of embracing AI within radiology. While AI algorithms have shown significant potential in improving diagnostic accuracy, their effectiveness is still under study. Understanding algorithm accuracy and addressing potential limitations and biases for reliable interpretations is essential. Furthermore, our research will compare AI utilization in US radiology departments and other countries, providing insights into AI's current status and prospects in radiology. By tackling the lack of understanding around AI amid information overload and concerns about job displacement and reduced human interaction, our study aims to demystify AI and facilitate its ethical and legal integration into US hospital infrastructures. There is an urgent need to improve the effectiveness and efficiency of clinical care, especially in radiology, where accurate and timely diagnoses are critical for patient outcomes. To achieve this goal, we can leverage AI 3 technology to help us better serve our communities and deliver high-quality care. Understanding that AI should not replace human interaction and care but rather serve as a valuable tool to assist radiologists in making accurate diagnoses and improving patient outcomes is essential. Ultimately, we aim to advance radiology practices by offering practical recommendations for integrating AI effectively while upholding ethical and legal standards, thus optimizing patient care and healthcare efficiency. Purpose of the Study This systematic review aims to provide clarity to radiology professionals regarding the current landscape of AI implementation in the United States and similar countries. Areas of focus include the effectiveness and accuracy of clinically used AI systems, the ethics and legalities of using AI in healthcare, and anticipated future developments. By exploring these dimensions, the study seeks to offer a comprehensive understanding of the role of AI within medical imaging today and its potential trajectory. With the shortage of radiologists and x-ray technologists entering the healthcare field, this study hopes to demonstrate the transformative potential of AI in their profession and dispel the fear of it replacing these professions in radiology practice. Methods The information for this study was systematically assembled through a comprehensive examination of peer-reviewed journal articles sourced from reputable databases such as Google Scholar and Weber State University's Stewart Library OneSearch. Additional general background information was obtained from professional online resources and websites. These sources were consulted between October 2023 and May 2024. A matrix of articles was organized 4 and included information on authors, source, subtopic, sample size, as well as strengths and weaknesses. All articles were published within the last 10 years and included information from developed countries comparable to the US. Sources were screened and exclusions were recorded via a flowchart (see Appendix A), per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines. Research Questions Q1. What emerging AI systems are currently developing, and what is their effectiveness? The rapidly evolving future of technology is at the forefront of our present world. Efficiency is the key to any advancement, and understanding what essential steps are needed to achieve that goal is fundamental to the future and understanding of AI. There have been limited studies about the effectiveness of AI in assessing strokes. Through these studies, one from 2017 and the other from 2023, we found the specificity of these studies has remained in the mid-90% range while the sensitivity of the software varies between 60-100%. Improving the variance of effectiveness requires future efforts. These results also support the need for an accredited neuroradiologist to review every scan and that AI software is not perfect, nor their replacement. Of course, human error is inevitable, but humans have the ability to adapt and think critically (i.e. accounting for possible imaging issues, drawing clinical conclusions, etc.). A computer does not have the same leniency. Q2. How is AI being adopted in healthcare systems of countries with socioeconomic contexts similar to the United States, particularly where implementation is advancing rapidly? The integration of AI into healthcare systems in nations with similar socioeconomic 5 backgrounds to the United States have experienced notable growth, particularly across Europe. Many European countries and other analogous nations have adopted AI-driven systems to prioritize critical tasks such as analyzing abnormal chest X-rays, lung cancer screening, head and neck radiotherapy, neuroimaging, and breast cancer screening. AI is also used to detect pulmonary emboli, parenchymal nodules, cerebral hemorrhage, and colonic polyps in some European nations. These algorithms have been beneficial in reducing workload and improving efficiency. Q3. How is AI utilized in the US? AI is currently used to improve the clinical workflow, most notably in the neurology and cardiology aspects of imaging. Cardiac imaging utilizing computed tomography (CT) guidance has been formatted for AI algorithms to assess cardiac functionality and vessel patency. CT is also the preferred imaging modality when assessing for a stroke. Head and neck perfusion imaging has advanced rapidly using algorithms developed by Viz.ai, which analyzes scan information and develops a preliminary report for the neurologist to review with an advanced evaluation in multi-functional formats. Q4. Is the United States falling behind in AI adoption, and are legal issues hindering progress? Liability issues have limited the use of AI in the United States healthcare system, as there is uncertainty as to who will be held responsible for any mishaps that may occur. While physicians can be liable for malpractice, hospitals can be guilty for vicarious liability. Similarly, software developers could be held responsible for injury resulting from poor design, manufacturing defects, or failure to warn about risks. The best solution to this issue depends on how much AI innovation one considers optimal and balanced with one's view on compensation 6 for injury. Liability directly affects the development and implementation of clinical algorithms. Increasing liability could discourage physicians, healthcare systems, and software developers from implementing AI algorithms. Nonetheless, it merits acknowledgment that radiologists responsible for issuing the conclusive report are liable for malpractice. Comprehensive legislation related to medical AI still needs to be established, which limits the role and growth of AI within the United States healthcare infrastructure. Q5. Is the integration of AI ethically sound, and what are the underlying issues? Are there potential breaches of HIPAA laws associated with integrating AI? Per ethical responsibility, patients have the right to know about their health status, diagnoses, treatment plans, test results, therapeutic status, cost, and health insurance. Informed consent involves appropriate communication between the patient and healthcare provider, which includes decision-making, documenting informed consent, and ethical disclosure. However, it is essential to consider whether patients have the right to be informed about AI involvement in their treatment process. Patients should be aware of the treatment plan, risks associated with screening and imaging, data capture anomalies, programming errors, data privacy, and access control. A data breach could undermine the validity of the healthcare system, exposing patients' sensitive information and putting healthcare professionals at risk of violating HIPAA laws. Q6. What does the future of AI look like? The future of AI in radiology appears promising. AI can enhance patient outcomes and take on repetitive tasks so radiologists can focus more on critical analysis. Incorporating AI into the imaging workflow can improve efficiency and minimize errors. Developers are creating tools such as AI-powered lung cancer and breast cancer detection systems, providing invaluable 7 support to radiologists and enhancing diagnostic capabilities. These advancements offer significant assets that elevate the overall field of radiology, empowering healthcare providers to deliver more accurate and timely diagnoses while improving patient outcomes. Nature of the Study Our study delves into the integration and impact of AI in radiology, focusing on its potential to enhance diagnostic accuracy, streamline workflow efficiency, reduce radiation and contrast use, improve patient care outcomes, and generate economic benefits. This research relies on empirical evidence and expert opinions, published articles, research papers, and reports as instruments for data collection and analysis. Employing a literature review methodology, it synthesizes findings from diverse sources related to AI applications in radiology. It analyzes data to evaluate the effectiveness and implications of AI-driven programs in enhancing radiological practices and patient care outcomes. Our mixed-methods approach allows for a comprehensive exploration of the multifaceted impact of AI in radiology, considering both quantitative metrics such as diagnostic accuracy rates and qualitative factors like patient care experiences and radiologist workflow improvements. Significance of the Study Our study aims to investigate the integration and impact of AI within hospital infrastructures in the United States compared to similar analogous nations. The study will provide insights into AI's potential for revolutionizing diagnostic accuracy, workflow efficiency, and patient care outcomes. It will also explore ethical boundaries and provide insight into maintaining HIPAA regulations. Additionally, the study will highlight AI's ability to address the challenges faced by radiologists in meeting increased workload demands, emphasizing the 8 potential for AI to streamline workflows while boosting productivity and reducing costs associated with redundant medical procedures. The findings will underscore the significant economic benefits of AI integration, including a substantial reduction in US healthcare spending. This study seeks to contribute to the field by offering a comprehensive overview of AI's transformative impact in radiology across various regions. It will address the critical need for improved diagnostic accuracy, workflow, and economic efficiency within the US healthcare system. The study will also compare the usage of medical AI in the US to other countries, as well as delve into the future growth of AI and its effects on radiology.. The study will also provide insights on the incorporation of AI within US hospital infrastructures while upholding ethical and legal laws. Definition of Key Terms AI - Artificial Intelligence. A branch of computer science focused on developing software to think and solve problems like humans. AUC - Area-under-the-curve often refers to the area under the Receiver Operating Characteristic (ROC) curve; it provides a single scalar value representing the classifier's performance, where a higher AUC value indicates better discrimination ability. CAD - Computer-aided diagnosis CE Mark - A self-declaration mark manufacturers use to prove compliance with European Union health and safety regulations. 9 Deep Learning - Subcategory of machine learning where AI programs automatically learn from prior data, becoming more accurate/efficient without further programming. Ground Truth - Often used to validate or train algorithms by providing a known correct answer or outcome for comparison. Machine Learning - Programming of computer software to recognize patterns and relationships within collected data. Neural Networks - Organization of computing layers that mimic the architecture of the human brain. ROC curves - Receiver Operating Characteristics, graphical representation that illustrates the diagnostic ability of a binary classifier system as its discrimination threshold is varied. Workflow Efficiency - The effectiveness and productivity of radiology processes. Diagnostic Accuracy - The precision and reliability of radiologic diagnosis. Patient Care Outcomes - The impact of radiologic interventions on patient health and wellbeing. HIPAA - Health Insurance Portability and Accountability Act is a federal law requiring national standards to protect sensitive patient health information from being disclosed. Summary This document delves into the current state of AI integration within medical imaging in the United States, aiming to provide a thorough understanding of its uses and challenges. The 10 exponential growth within the global medical AI market requires enhanced clarity and comprehension, helping lessen concerns regarding job displacement while promoting legislative support. With radiologists facing immense workload demands and time constraints, this study emphasizes AI's potential to refine workflow efficiency, enhance diagnostic accuracy, and improve patient care outcomes. By exploring ethical, legal, and technological elements, this document aims to educate readers on the challenging nature of AI implementation in healthcare and the significance of future AI advancement within radiology. 11 Chapter 2: Literature Review The following literature review analyzes the current state of artificial intelligence (AI) in the healthcare systems of developed countries. After introducing general background information, this literature review delves into the documented potential of AI and how it’s currently being used both in the US and internationally. Limiting factors of legislation and societal/ethical concerns are reviewed, and finally, areas for future research and improvement are discussed. This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines. Documentation This study synthesizes information gathered through scholarly databases, including Weber State University Stewart Library OneSearch and Google Scholar between October 2023 and May 2024. A systematic review was performed utilizing a “key concept” research strategy. Primary resources consist of published peer-reviewed articles. General web and news articles were also referenced for context and general background information. Introduction and Background Artificial intelligence (AI) refers to the design of computer applications that think, learn, and problem-solve like humans. With the help of cutting-edge processing power and leaps in data storage, AI has become a functional tool with applications emerging in almost every industry, including health care. While modern healthcare practices have significantly relied on computers for quite some time, AI is opening doors for better workflow and, ultimately, better patient outcomes. 12 The field of radiology has become a hotspot for AI implementation. Since the development of digital imaging techniques, computers have been used to immediately acquire and manipulate radiology exams. After an examination, radiologists review digital images and directly dictate reports into the patient's electronic medical record. Other than the initial patient exam or procedure, most radiology tasks occur within a computer, making radiology the perfect sector in healthcare for AI growth. One significant advancement in AI technology is the development of machine learning. Machine learning is a subcategory of AI computer science that uses algorithms to recognize patterns and relationships in collected data for a specific task. After programmers program and train an algorithm, they expose it to image patterns, rendering it capable of performing tasks on par with those executed by a well-trained, experienced radiologist (Wang & Summers, 2012). Beyond machine learning lies an even more advanced system called deep learning. Deep learning is a revolutionary AI architecture with multiple computing layers organized in a "neural network" inspired by the human brain. These systems autonomously adapt and become increasingly more efficient at identifying complex patterns, shapes, organs, and pathology over time as they collect more and more data (Chartrand, 2017). AI systems can speed up or replace time-consuming tasks by pre-examining complex imaging patterns and automatically marking them for the radiologist, making for a quicker and more confident diagnosis. Thus, stakeholders are implementing applications across all imaging modalities. While the idea of artificial intelligence sounds ominous, it is shaping up to be an undeniably powerful tool capable of changing healthcare for the better. Radiology practices and hospitals worldwide are already taking advantage of the diagnostic confidence, workflow efficiency, and accuracy offered by these sophisticated tools. 13 Accuracy and Efficiency AI-driven programs offer promising opportunities to enhance diagnostic accuracy, streamline workflow efficiency, and improve patient care outcomes. The objective of AI is to construct systems that are capable of learning, reasoning, and adapting through the use of data and feedback. Radiologists can seamlessly integrate AI throughout various diagnostic imaging processes, including data acquisition, reconstruction, analysis, and reporting (Boeken, 2023). With AI-driven tools, radiologists can strategically manage their time, dedicating more attention to interpreting intricate cases and performing interventions. AI can offer invaluable tools to streamline workflows for radiologists, helping to transform the efficiency and effectiveness of diagnostic imaging. By automating routine tasks such as preliminary reads, scheduling patient appointments, and patient no-show predictions, AI systems can significantly impact the time and effort required for image interpretation and patient care. The study by van Leeuwen et al. (2021a), demonstrated that an AI model could predict which patients had the highest risk of missing their appointments. This predictive capability allowed for targeted reminder calls that reduced the no-show rate from 19.3%-15.9%. Van Leeuwen et al. (2021a) also noted that these AI solutions primarily address operational challenges like patient management rather than direct diagnosis, which means they are subject to fewer regulatory hurdles and pose lower risks. Efficient workflow management remains essential amid rising global healthcare costs. Barat et al. (2023) highlighted that the high pace at which radiologists work, interpreting one image every 3-4 seconds during an 8-hour shift, inevitably leads to errors under such intense conditions. This reality underscores the need for AI to enhance both efficacy and efficiency in clinical practice. By leveraging AI technology in appointment scheduling and prediction, radiology departments can streamline workflow processes, minimize 14 appointment delays, and enhance operational efficiency, ultimately improving patient care experiences and outcomes. Enhancing diagnostic accuracy and image analysis is a central objective of many AI solutions in radiology. The integration of AI offers radiologists promising capabilities in pathological identification through advanced pattern recognition and data analysis algorithms. Notably, AI aids in detecting subtle features and abnormalities, as evidenced by Tam et al. (2021). This research focused on chest radiograph reads for lung cancer detection and demonstrated a 60% reduction in missed diagnoses when using AI assistance. The integration of AI led to an overall improvement in accuracy and sensitivity for tumor identification among radiologists, with average scores increasing by +3.67% and +13.33%, respectively. Moreover, the integration of AI notably reduced false-negative cases, where cancer findings were missed, by between 15 and 40 cases. These findings highlight the transformative impact of AI in elevating diagnostic capabilities and, consequently, improving patient care outcomes in radiology. AI algorithms, like Computer-Aided Detection (CAD) systems, use machine learning techniques to identify image patterns and features, similar to how CAD algorithms detect specific patterns or abnormalities like suspicious lesions in mammograms or lung nodules in chest X-rays. This assists radiologists in making accurate diagnoses and refining their clinical decision-making process. Significantly, this approach not only enhances diagnostic accuracy but also can significantly impact read times. As described in the literature by van Leeuwen et al. (2021a), A study performed by Martini et al. found that vessel suppression on CT thorax 15 imaging resulted in a 21% reduction in reading time for the detection of pulmonary metastasis. Moreover, the automated quantification of nodules, brain volumes, or other tissues, for example, might mitigate some tedious manual work that is part of a radiologist's job, along with the large interrater variability inherent to these tasks (p. 2089). As this research shows, AI algorithms are pivotal when streamlining routine tasks like image segmentation, annotation, and preliminary analysis by swiftly processing large amounts of imaging data and pinpointing regions of interest. Furthermore, AI algorithms continuously evolve and refine their performance on the basis of feedback and new data, enhancing the efficiency and accuracy of image interpretation and enabling radiologists to make faster, more informed decisions. AI holds immense potential in reducing radiation and contrast use in medical imaging, thus enhancing patient safety and minimizing associated risks. Using advanced algorithms and machine learning techniques, AI can optimize imaging protocols to achieve high-quality diagnostic images using lower radiation doses. In addition, AI-powered image reconstruction methods can generate high-fidelity images from lower-dose scans, mitigating the need for excessive radiation exposure. Van Leeuwen et al. (2023) described software such as image reconstruction and post-processing as technology that facilitates good image quality with a lower dose or even no dose. A study from Belgium demonstrated this by using commercial AI software to synthesize a CT from MR to assess lesions in the sacroiliac joints and diagnose spondyloarthritis, maintaining diagnostic accuracy and potentially rendering CT redundant. Similarly, AI algorithms can optimize contrast administration protocols by tailoring 16 contrast doses to individual patient characteristics and clinical indications, thereby minimizing unnecessary contrast use and associated adverse effects. Through continuous monitoring and feedback, AI systems assist radiologists and technologists in maintaining radiation exposure within safe limits while preserving image quality. AI's ability to optimize imaging protocols, reconstruct images, and optimize contrast administration significantly reduces radiation and contrast use in medical imaging, aligning with maximizing patient benefit while minimizing potential harm. The use of AI in radiology also offers significant economic benefits to radiologists and healthcare systems. An article reported by the World Economic Forum (2018) estimated that the top 15 countries waste an average of $1,100 to $1,700 per person annually due to preventable inefficiencies within their systems, such as unnecessary treatments and improper care procedures. Simplifying workflows through task automation, such as image analysis, allows radiologists to concentrate on more intricate cases, boosting productivity and cutting labor expenses. AI can also utilize its personalized treatment strategies, drawn from thorough patient data analysis, to minimize unnecessary procedures and reduce costs linked to overtreatment. This, combined with AI-driven decision support systems, reduces errors and misdiagnoses, curbing the need for repeat imaging studies and associated costs. A 2023 report from the National Bureau of Economic Research suggests that the integration of AI could reduce US healthcare spending by 5 to 10%, saving overall US healthcare spending. These results further support the initial investment in AI technologies, highlighting the long-term benefits of improved efficiency, accuracy, and patient outcomes, ultimately leading to substantial cost savings within radiology departments. Uses Overseas 17 Advancements in medical technology continue to evolve globally. Within the United Kingdom's National Health Service (NHS), several trusts have adopted advanced AI platforms, including Behold.ai's red dot®, which prioritizes abnormal chest X-rays for lung cancer detection and neuroimaging for stroke diagnosis, and Mammography Intelligent Assessment (MIA), a specialized tool exclusively utilized as a second reader for breast cancer screening (see Appendix B). The European and Chinese medical communities have collaborated with companies like Infervision, DeepMind, and Nvidia to aid critical areas such as lung cancer screening, head and neck cancer radiotherapy, neuroimaging, and breast cancer detection (Mudgal et al., 2020). Respondents to a survey of European radiologists reported various uses of AI, the most common being the detection or marking of specific findings like pulmonary emboli, parenchymal nodules, cerebral hemorrhage, and colonic polyps. Moreover, AI extracts quantitative imaging biomarkers to objectively measure disease progression and treatment response in European countries. Prioritization algorithms were found to be beneficial, reducing the workload and improving overall efficiency (Becker et al., 2022). One example of a commonly used prioritization algorithm used in the UK is Behold.ai’s red dot®CXR solution. The red dot®CXR solution uses AI to streamline lung cancer diagnostics, reducing radiologists' workload by 15% and prioritizing urgent cases for faster CT imaging (see Appendix C). Its heatmap highlights suspicious areas, aiding radiologists, and shortens the timeline from chest Xray to CT from 7 to 3 days, improving efficiency and care (see Appendix D). In Europe, 100 CE-marked AI products are commercially available. There is a significant focus on neuroradiology, chest radiology, breast radiology, and musculoskeletal radiology. However, only 36 CE-marked products have peer-reviewed evidence on the efficacy of AI software, emphasizing the need for further research on the clinical impact of AI solutions. 18 Approximately two-thirds of the software in use has been on the market in the past few years, which was significant during the recent COVID-19 outbreak (van Leeuwen et al., 2021b). During the COVID-19 pandemic, Siemens' CE-label approved CT Pneumonia Analysis2 was employed in European countries to identify and quantify lung abnormalities associated with COVID-19 pneumonia (Siemens AG, 2020). By providing detailed metrics, such as Percentage of Opacity (PO) and Lung Severity Score (LSS), the tool supported radiologists in assessing disease severity and monitoring progression, enhancing diagnostic accuracy and patient management in overwhelmed healthcare systems (see Appendix E). By examining specific cases, the implementation of AI has proven invaluable in predicting the clinical outcomes of patients with COVID-19. A retrospective study from Italy used deep learning to analyze chest radiographs, demonstrating AI's potential for identifying critical cases early on (Mushtaq et al., 2021). According to a study from the Netherlands, the diversity of AI products implemented increased fivefold, showing a trend toward increased acceptance and integration of AI technology. The survey also presented the growing adoption of AI, primarily in academic and teaching hospitals (van Leeuwen et al., 2023). The use of AI in radiology is increasing in Europe and other countries. It shows promising signs of improved detection and accuracy in diagnosing many pathologies, streamlining workflows, and enhancing patient outcomes. There are currently 100 commercially available AI products with CE marking, notably those concentrating on neuroradiology, chest radiology, breast radiology, and musculoskeletal radiology. Countries outside the United States use AI in critical areas such as lung cancer screening, head and neck cancer radiotherapy, neuroimaging, and breast cancer detection. These other countries and the United States also face challenges such as regulatory frameworks and evidence-based assessments. As the field evolves, 19 collaboration between clinicians, technology developers, and regulatory bodies will ensure AI's responsible and effective integration in foreign radiology departments. Uses in the USA In the current landscape of X-ray procedures, understanding clinical workflow is crucial, showcasing AI's already-established role in significantly boosting the efficiency of imaging technologists. In 2022, the FDA authorized 91 AI- or machine-learning-enabled medical devices for clinical use, with over 500 such devices approved since 1995, displaying a significant presence of AI tools in radiology. "Of the 521 submissions the FDA has authorized to date, three-quarters have been in radiology, and 11% have been in cardiology" (Reuter, 2022). [CV4] [LC5] US healthcare providers have widely integrated AI into imaging modalities and use it extensively to assess various aspects of imaging-related healthcare. For example, cardiac imaging performed using CT assists cardiologists in assessing heart function, patent vessels, and functionality of aspects of the heart. In addition, CT is utilized by vascular surgeons to evaluate proper flow in a patient’s extremities. If an occlusion is present, AI can help target the affected area. The seamless integration of AI algorithms into imaging systems has led to the immediate benefits of real-time quality control alerts and automated image rotation. A prime example is GE Healthcare's groundbreaking Quality Care Suite of X-ray AI solutions, which actively operates in parallel with image acquisition. This reduces image quality errors and enhances the overall efficiency of examinations. Using this advanced software, AI performs automated protocol checks, ensures precise usage, and automatically rotates images for most X-ray exams. This ensures the swift and accurate transmission of all images, appropriately labeled according to their 20 corresponding protocols, which is especially crucial during multiple examination imaging scenarios.The real-time support provided by AI proves to be an indispensable asset, enabling technologists to promptly assess the readiness of images for radiologist review or identify the need for potential retakes before the patient concludes their examination. This demonstrates the tangible and current impact of AI on elevating workflow efficiency in X-ray procedures (GE Healthcare, 2023). Moreover, AI is already revolutionizing post-scanning image reconstruction, representing a significant benefit for imaging technologists. Industry leaders like GE Healthcare have invested considerable efforts in image reconstruction stages, leveraging deep-learning algorithms to increase signal-to-noise ratios. This has resulted in 30% faster imaging, higher quality, and improved resolution (Cowen, 2023). The compatibility of this tool with most companies' installed CT and MRI systems exemplifies how AI is actively contributing to increased productivity in daily patient scans. The cost-effectiveness of this approach has emphasized that integrating AI into existing systems is not only about improving patient outcomes and safety but also about driving significant gains in productivity without the need for extensive infrastructure changes. Simultaneously, Cowen (2023) also boasts about how other industry leaders like Philips, Siemens, and Canon are incorporating their deep-learning algorithms into their imaging systems, collectively contributing to an ongoing paradigm shift in image reconstruction and workflow efficiency in the United States. Radiologists have strategically integrated diverse AI algorithms into their daily practices, marking a transformative collaboration that found its roots in the 1990s. During this period, the advent of CAD systems led to a significant breakthrough by providing invaluable support to radiologists, particularly in the secondary interpretation of mammogram imaging. The evolution 21 of CAD algorithms employed in mammogram interpretation has been profound, attaining a level of sophistication that qualifies for Medicare and Medicaid insurance reimbursement. This not only underscores the instrumental role of AI in contemporary radiology practice but also reflects the maturation of AI technologies over the years. The collaborative journey between radiologists and AI began with the FDA approval of the first commercial CAD system for screening mammography in 1998 (Cowen, 2023). Since this milestone, subsequent CAD systems for mammography have also gained FDA approval, with the extension of approval to CAD systems for digital mammography. This regulatory validation has paved the way for the widespread clinical adoption of numerous CAD systems used in screening screen films and digital mammography in the United States and internationally. The expansive use of these systems underscores their clinical efficacy and acceptance, further emphasizing their integral role in modern medical imaging practices. AI actively contributes to bone health assessments by identifying bone mineral density (BMD), particularly in interpreting dual-energy X-ray absorptiometry (DEXA) scan images. This intelligent software adeptly discerns variations in bone density, providing radiologists with precise density scores tailored to individual patients. Incorporating patient demographics, such as age and ethnicity, into the AI system allows for accurate categorization of T-scores and Z-scores within specific demographic groups. This personalized approach ensures a delicate evaluation of bone health, considering the unique characteristics of each patient. LiberaBMD, an automated BMD reporting system that uses raw data from a Hologic Delphi series DXA scanner, is a noteworthy illustration of AI's impact (see Appendix F). This system underscores the efficiency of AI by significantly reducing the time spent on report generation and achieving impeccable accuracy with a 100% accuracy rate in T and Z scores within automatically generated reports 22 (Tsai et al., 2016). The seamless integration into electronic medical records (EMR) systems streamlines radiologists' workflows and enhances assessment precision, minimizing the risks associated with manual data entry errors. The application of AI in DEXA exemplifies its efficacy in error prevention, ensuring precise bone health assessments, and contributing to the advancement of patient care in the radiology domain. The CardIQ Suite in CT, another groundbreaking AI application with FDA approval, is currently in widespread use across the United States. Developed by GE Healthcare, this FDAapproved noninvasive software harnesses AI in multiple pivotal aspects to optimize its functionality. Tailored for analyzing cardiovascular anatomy and pathology through 2D or 3D CT cardiac data, the software is crucial for visualizing vessels, measuring chamber mobility, and diagnosing various cardiovascular diseases, including coronary artery disease (see Appendix G). One standout feature of AI integration within CardIQ Suite is its calcium scoring functionality, which automatically identifies and labels calcifications within coronary arteries using sophisticated deep-learning algorithms. This AI-driven capability significantly improves workflow efficiency by automating the calculation of total and per-territory calcium scores, which are essential for monitoring disease progression or regression. The application demonstrated over 90% consistency in concordance with the Coronary Artery Calcium Data and Reporting System (CAC-DRS) classification grouping, over 95% accuracy in identifying the presence of coronary artery calcifications, and over 90% precision in labeling coronary artery territories (GE Healthcare, n.d.). Moreover, the software offers manual editing options for identified lesions and facilitates precise quantitative assessment of cardiac structures. Overall, the seamless integration of AI into the CardIQ Suite enhances diagnostic accuracy and aids in treatment planning, ultimately contributing to improved patient outcomes in cardiovascular care. 23 In computed tomography (CT), AI has been widely used across the United States since the creation of Viz.ai in 2016 by two doctors with the intent of achieving efficiency and accuracy for large vessel occlusions (LVO) during stroke imaging. Since the creation of Viz.ai, many researchers have conducted experiments to test the accuracy and prompt detection of the software, and the results have provided intriguing insights into the future of AI in radiology. A study conducted between March 2016 and November 2017 demonstrated that out of 650 imaging studies assessed by the Viz.ai software, the sensitivity was 82% with a specificity of 94% with 31 false positives and 23 false negatives (Chatterjee et al., n.d.) The algorithms used in this software are based on a convolutional neural network that follows the path of anterior circulation LVOs (see Appendix H). These results prove that although AI can help provide faster results, it may need to be more efficient. This study also helped to identify that as time went on, the algorithm did not adapt to become more successful, enhancing the rationale of radiologists' necessity in diagnosing neuroradiologic findings. Another study was conducted using Viz.ai software from January to December 2021, which included assessing LVOs and detecting internal carotid artery (ICA) or middle cerebral artery first segment (MCA-M1) occlusion. In this study, neuroradiologists assessed and compared the results of studies conducted using AI algorithms. The criteria for inclusion in the study were 3851 patients with an ICA or MCA-M1 occlusion, as determined by the NRs reviewing the images. The data are shown below for review in Table 1[LC6] . Table 1 Viz.ai Software analysis in assessing Large Vessel and Internal Carotid Artery or middle cerebral artery first segment occlusions Month January True Negatives True Positives False Negatives False Positives Sensitivity (%) Specificity (%) (n) (n) (n) (n) 262 12 5 15 70.6 94.6 24 Month True Negatives True Positives False Negatives False Positives Sensitivity (%) Specificity (%) (n) (n) (n) (n) February March April May June July August September October 290 319 290 342 310 303 280 279 263 13 15 13 23 12 12 16 11 13 9 4 3 8 0 0 4 6 4 10 10 5 7 10 8 9 9 12 59.1 78.9 81.3 74.2 100 100 80 64.7 76.5 96.7 97 98.3 98 96.9 97.4 96.9 96.9 95.6 November 286 20 1 6 95.2 97.9 December 298 12 4 8 75 97.4 Total 3522 172 48 109 78.2 97 Note. Evaluation of artificial Intelligence-powered Identification of large vessel occlusions in a comprehensive stroke center Algorithm Assessment of collected data Legalities/Ethics As AI gains speed within the medical field in European countries, the following question arises: What is halting medical professionals in the United States from adopting these likeminded studies? When delving deeper into recent studies, The Milbank Quarterly (Maliha, 2021), one reservation among medical professionals raised concerns about legality and ethics. However, some medical imaging techniques have adopted AI and are used daily within the United States hospital infrastructures, such as DEXA scans, mammography, and CT stroke perfusion, to name a few prevalent studies. The next question is, "Why is artificial intelligence not being used more?" The liability issue has limited the use of AI in the United States healthcare system. The significant uncertainty is who will be held responsible for any mishaps that may occur. While physicians can be liable for malpractice, hospitals can be liable for vicarious liability. Similarly, software developers could be held responsible for injuries resulting from poor design, 25 manufacturing defects, or failure to warn about risks. The best solution to this issue depends on how one defines optimal AI innovation in balance with their view on compensation for injury. Liability directly affects the development and implementation of clinical algorithms. Increasing liability could discourage physicians, healthcare systems, and software developers from implementing AI algorithms. Radiologists who issue conclusive reports are liable for malpractice. The laws related to AI still need to be established, limiting AI's role and growth within the healthcare infrastructure of the United States. Before integrating AI with healthcare systems, the healthcare industry must consider several ethical challenges during artificial intelligence's revolutionizing impact. These include informed consent, privacy and data protection, medical consultation, empathy, and sympathy. Therefore, healthcare practitioners and specialists must prioritize the four medical ethics principles - autonomy, beneficence, nonmaleficence, and justice - in all aspects of healthcare (Farhud, 2021). AI in the healthcare industry analyzes consumer health data, enhances diagnoses, and accelerates health research activities. However, for AI to improve patient outcomes, access to patient demographics for programming and machine learning is required. The Genetic Information Non-discrimination Act is a US-based organization that prohibits employers from making discriminatory decisions based on individuals' genetic health information. Although using AI can have obvious benefits, ensuring the safety of patients' data is still a significant concern when using AI tools. Per ethical responsibility, patients have the right to know about their health status, diagnoses, treatment plans, test results, therapeutic status, cost, and health insurance. Informed 26 consent involves appropriate communication between the patient and healthcare provider, which includes decision-making, documenting informed consent, and ethical disclosure. However, it is essential to consider whether patients have the right to be informed about AI involvement in the treatment process. Patients should be aware of the treatment plan, risks associated with screening and imaging, data capture anomalies, programming errors, data privacy, and access control. A data breach could undermine the validity of the healthcare system, exposing patients' sensitive information and put healthcare professionals at risk of violating HIPAA laws. Integrating artificial intelligence into various aspects of the medical field may harm human emotions. Although AI has demonstrated several positive patient outcomes, there is concern that it could diminish human-to-human interaction. Patients may experience less empathy and kindness as their care becomes less human and more automated. This could notably affect particular patient groups, such as children, psychiatric patients, and those in obstetrics and gynecology (OBGYN). The rapid progress of artificial intelligence (AI) in clinical and biomedical fields is a promising approach that can assist healthcare professionals. However, with advancements come new ethical considerations to protect patients. To address these concerns, experts should prioritize ethical boundaries while striving for improved patient outcomes. The future of AI in healthcare is inevitable, and achieving a balance between progress and moral considerations should be a collaborative effort. Future of AI While interest and research in AI in radiology have surged, addressing numerous substantial challenges is necessary before the broad implementation and integration of AI into 27 clinical practice become feasible. Key issues include ethical development and use of AI in healthcare, proper validation of each AI algorithm, the establishment of efficient data-sharing mechanisms, and the creation of educational resources on AI for both practicing radiologists and radiology trainees (Dewey, 2020). The future is bright for AI, enhancing the impact of radiology on patient outcomes and elevating the overall importance of the field. Instead of focusing exclusively on memorizing patterns and recognizing differential diagnoses, radiologists will shift toward becoming data or information managers. Their role will encompass integrating and interpreting imaging data, pathology findings, and genomic information for the patient's clinical care team. An incorporated system like this will enable swift on-site point-of-care diagnostics, eliminating the current continuous procedure that requires multiple appointments spread across several days and weeks. As AI takes on repetitive and time-consuming tasks, it has the potential to amplify the purpose and satisfaction derived by radiologists (Dewey, 2020). Integrating an AI component seamlessly into the imaging workflow will enhance efficiency, minimize errors, and achieve objectives with minimal manual effort. Radiologists will benefit from pre-screened images and identified features, prompting significant efforts and policies to promote technological advancements in AI for medical imaging. Nearly all radiology tasks that rely on images hinge on assessing radiographic characteristics. These characteristics are crucial in clinical tasks such as disease detection, characterization, and monitoring (Hosney, 2018). Germany is expected to lead in using AI in healthcare and be a significant innovator in the artificial intelligence industry. The German Research Foundation recently launched a new 28 artificial intelligence program that prioritizes medical imaging. The program's primary goal is to assess medical imaging data and use it to obtain new image information for diagnostic evaluation. The program is expected to expand the search for complex data with precise patterns, using high-throughput imaging and post-processing. This data will be used to create a method to search for comparable patient cases with suitable therapies (MTR Consult, 2019). In addition, one of the largest university hospitals in Europe is developing algorithms powered by AI for imaging analysis, such as tumor and trauma detection. A tool they are creating is an artificial intelligence-powered tool to detect lung cancer on CT scans. This health system has also created a system to accurately detect breast cancer in mammograms. These innovations and dedication would benefit the United States in incorporating them into future artificial intelligence models. Seeing one of the leaders in medical imaging making such successful strides creates an achievable standard for radiologists who may feel skeptical about using AI in medical imaging (Patel, 2023). Summary After extensive review, it is clear that healthcare systems across the globe are already benefiting from AI assistance. Approximately 100 AI products are already approved and in-use within the European Union (Leeuwen et al., 2022b), and 91 AI products received FDA approval within the US in 2022 alone (Reuter, 2022). Not only is AI improving diagnostic accuracy, evident by a shocking 60% decrease in missed lung cancer reads in one study (Tam et al., 2021), it is equally powerful in improving non-clinical tasks, such as scheduling and worklist prioritization. While implementation is proving to be valuable, liability concerns appear to slow down integration, and societal fears may be causing a resistance to change. Nevertheless, the 29 benefits are undeniable. With continued research, governmental support, and ethical persistence, AI assistance will significantly change the quality of healthcare. 30 Chapter 3: Research Method For years, advancements in radiology and computer technology have been closely linked. However, the rapid evolution of Artificial Intelligence (AI) is transforming this relationship. A recent article in Diagnostic and Interventional Imaging highlighted that radiologists are now required to review an image every 3-4 seconds to keep up with their workload demands (Barat et al., 2023). The emergence of innovative AI applications offers promising solutions for managing these challenging workloads. With numerous medical journals, magazines, and biomedical manufacturers promoting the latest AI breakthroughs, staying current with these developments can be overwhelming. This study aims to provide a comprehensive examination of the current applications, effectiveness, legal and ethical issues, and future prospects of AI in radiology. Integrating AI into radiology departments is a critical area of focus in contemporary U.S. healthcare, especially when compared to similar nations. Our research is centered on exploring AI's potential to enhance diagnostic accuracy, streamline workflows, and improve patient care outcomes. We will examine the ethical implications of AI implementation, ensure compliance with regulations like HIPAA, and address concerns related to patient privacy and data security. Our study aims to highlight the challenges radiologists face with increasing workloads and explore how AI can optimize workflows, boost productivity, and reduce costs associated with unnecessary medical interventions. By emphasizing the economic advantages of AI integration, particularly the anticipated reduction in U.S. healthcare expenditures, we underscore the need for AI adoption in radiology. While AI algorithms have demonstrated promise in enhancing diagnostic accuracy, their effectiveness remains under evaluation. It is crucial to 31 understand algorithm accuracy and address any potential limitations and biases to ensure reliable interpretations. Additionally, our research will compare AI use in U.S. radiology departments with its application in other countries, offering insights into AI's current state and future potential in radiology. By addressing the challenges of information overload and concerns about job displacement and reduced human interaction, our study seeks to clarify AI’s role and support its ethical and legal integration into U.S. hospitals. Improving the effectiveness and efficiency of clinical care, particularly in radiology where timely and accurate diagnoses are vital, is essential. AI should not replace human interaction but should be a valuable tool to assist radiologists in making precise diagnoses and enhancing patient outcomes. Ultimately, our goal is to advance radiology practices by providing practical recommendations for effective AI integration while adhering to ethical and legal standards, thus optimizing patient care and healthcare efficiency. Research Questions Q1. What emerging AI systems are currently developing, and what is their effectiveness? The rapidly evolving future of technology is at the forefront of our present world. Efficiency is always the key to any advancement, and understanding what essential steps are needed to achieve that goal is fundamental to the future and understanding AI. There have been limited studies about the effectiveness of AI in assessing strokes. Through these studies, one from 2017 and the other from 2023, the specificity of these studies has remained in the mid-90% range while the sensitivity of the software varies between 60-100%. Improving the variance of effectiveness requires future efforts. It also solidifies the need for an accredited neuroradiologist (NR) to review these images after being analyzed by the AI software. The need for a radiologist to review the images sets back the goal of AI taking over and being able to detect things reliably 32 on its own. Of course, radiologists also make mistakes, but it is easier for humans to learn new ways of looking at things and account for possible imaging issues. In contrast, a computer does not have that leniency. The study described in this article addresses the lack of adaptation and what it means for the future of AI. Q2. How is AI being adopted in healthcare systems of countries with socioeconomic contexts similar to the United States, particularly where implementation is advancing rapidly? The integration of AI into healthcare systems in nations with similar socioeconomic backgrounds has experienced notable growth, particularly across Europe. Many European countries and other analogous nations have adopted AI-driven systems to prioritize critical tasks such as analyzing abnormal chest X-rays, lung cancer screening, head and neck radiotherapy, neuroimaging, and breast cancer screening. They also use the technology to detect pulmonary emboli, parenchymal nodules, cerebral hemorrhage, and colonic polyps. These algorithms have been beneficial in reducing workload and improving efficiency. Q3. How is AI utilized in the US? AI is currently used in specific imaging areas to improve the clinical workflow, mainly in the neurology and cardiology aspects of imaging. Cardiac imaging using computed tomography (CT) guidance has been formatted to AI algorithms in assessing function, patent vessels, and functionality concerning the heart. CT is also the preferred imaging modality when assessing for a stroke. Perfusion imaging has advanced rapidly using algorithms developed by Viz.ai, which analyzes the collected images and data and develops a preliminary report to be sent back to neurologists with an advanced evaluation in multi-functional formats. Q4. Is the United States falling behind in AI adoption, and are legal issues hindering progress? The liability issue has limited the use of AI in the United States healthcare system. The 33 significant uncertainty is who will be held responsible for any mishaps that may occur. While physicians can be liable for malpractice, hospitals can be liable for vicarious liability. Similarly, software developers could be held responsible for injury resulting from poor design, manufacturing defects, or failure to warn about risks. The best solution to this issue depends on how much AI innovation one considers optimal and balanced with one's view on compensation for injury. Liability directly affects the development and implementation of clinical algorithms. Increasing liability could discourage physicians, healthcare systems, and software developers from implementing AI algorithms. Nonetheless, it merits acknowledgment that radiologists responsible for issuing the conclusive report are liable for malpractice. The laws related to AI still need to be established, and this limits the role and growth of AI within the United States healthcare infrastructure. Q5. Is the integration of AI ethically sound, and what are the underlying issues? Are there potential breaches of HIPAA laws associated with integrating AI? Per ethical responsibility, patients have the right to know about their health status, diagnoses, treatment plans, test results, therapeutic status, cost, and health insurance. Informed consent involves appropriate communication between the patient and healthcare provider, which includes decision-making, documenting informed consent, and ethical disclosure. However, it is essential to consider whether patients have the right to be informed about AI involvement in their treatment process. Patients should be aware of the treatment plan, risks associated with screening and imaging, data capture anomalies, programming errors, data privacy, and access control. A data breach could undermine the validity of the healthcare system, exposing patients' sensitive information and putting healthcare professionals at risk of violating HIPAA laws. Q6. What does the future of AI look like? 34 The future of AI in radiology appears promising. AI can enhance patient outcomes and take on repetitive tasks so radiologists can focus more on critical analysis. Incorporating AI into the imaging workflow can improve efficiency and minimize errors. Developers are creating tools such as AI-powered lung cancer and breast cancer detection systems, providing invaluable support to radiologists and enhancing diagnostic capabilities. These advancements offer significant assets that elevate the overall field of radiology, empowering healthcare providers to deliver more accurate and timely diagnoses while improving patient outcomes. Research Methods and Design(s) The following literature review analyzes the current state of artificial intelligence (AI) within developed countries’ healthcare systems. The study, conducted in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines, aimed to enhance patient care and satisfaction through the integration of artificial intelligence while safeguarding jobs within the healthcare sector. Data was sourced from thirdparty and peer-reviewed literature, focusing primarily on well-developed nations, including the United States, to provide a comprehensive understanding of how comparable countries are integrating AI. Data collection was carried out using various databases, including Google, Google Scholar, and the Stewart Library at Weber State University, with an emphasis on keywords such as "artificial intelligence," "ethics," "liability," "implementation," and "patient benefits." The findings underscore the importance of establishing a liability framework that equitably distributes responsibility across the healthcare ecosystem, promotes AI’s safe evolution and integration, and mitigates undue burdens on frontline clinicians and physicians. Population 35 As this literature review does not involve a research study, we do not have a typical population with participants. Instead, we will be gathering information from third-party and peerreviewed sources, primarily focusing on well-developed countries, including the United States. This approach will help us better understand how countries like ours are implementing AI. We have deliberately excluded studies from lesser-developed countries as the dynamics of AI implementation may differ significantly from those in the United States. Additionally, we will be cautious when using information from major A.I. manufacturers to avoid any potential bias. Sample In order to collect our data, we will be utilizing a range of databases, including Google, Google Scholar, and Weber State University Stewart Library. Our primary source of information will be peer-reviewed articles, which can be filtered on the databases we use. We will be focusing on specific keywords such as artificial intelligence, ethics, liability, implementation, and patient benefits to ensure that we obtain the most relevant information. Data Collection, Processing, and Analysis We have initiated collecting data by conducting preliminary research to narrow down our topic of interest. We have narrowed down the databases that are relevant to our needs and will continue to refine the keywords used in our searches, applying filters to find the most appropriate articles. Our chosen sources will be reviewed, evaluated, and organized into specific categories that align with our thesis. Once we have selected and evaluated the articles, we will synthesize the information to draw connections between the data, identify trends, and provide answers to our research questions. 36 As part of our research, we will be organizing peer-reviewed and third-party articles according to concepts related to our research questions. We will use an Excel spreadsheet with color coding to identify articles for each variable easily. The spreadsheet will contain information about the variables found in the article, quotes that can be used, and limitations of the research. It will also include information about the authors and publication for reference purposes. Since this information will be added to an online spreadsheet, as a group, we can cross-check information provided by others. Based on the extensive data we have reviewed from the group, we will be able to crossexamine the sources we have selected to verify their content validity. The authors or publishing groups of the respective articles will determine the reliability of each source. Additionally, crossreferencing sources will help to confirm the reliability of each source in relation to the others. Assumptions The following assumptions are key to our research conducted on the application of artificial intelligence in radiology. It is assumed that data used in this study is from legitimate and reputable sources. The study assumes that the AI models developed and referenced were created by companies with a goal to make this technology affordable, accessible and accurate. It is assumed that algorithms have been developed and tested to enhance accuracy and workflow. Limitations Outlined below are the limitations and considerations for the literature review. The primary limitation is the lack of literature on the use of AI in the United States, which reflects its relatively limited adoption. Additionally, we acknowledge the potential variation in research environments and subjects across different countries, which prompts a comparative analysis with 37 other well-developed nations. To address potential biases, we adopt a cautious approach regarding the origin of data by advocating for third-party studies and emphasizing the importance of avoiding data directly collected by AI software developers. We also acknowledge the historical limitation due to AI's recent development, which prompts the inclusion of findings from older AI applications in radiology, such as mammography AI. We recognize the rapid evolution of AI technology and, therefore, emphasize the need for up-to-date studies while cautioning against reliance on dated articles. These considerations aim to ensure that the literature selected for the review is relevant, reliable, and impartial. Delimitations This literature review focuses on AI integration in radiology within healthcare systems comparable to that of the United States. The principal delimitation involves including only peerreviewed articles and authoritative sources from developed nations with advanced healthcare infrastructures, such as the United Kingdom and Germany, and select European and Asian countries. These nations were chosen due to their similarity to the U.S. in healthcare systems, enabling meaningful insights into AI adoption and challenges within comparable socioeconomic and regulatory contexts. This review deliberately excludes studies from less-developed countries to ensure relevance, as economic limitations, disparities in healthcare infrastructure, and differing regulatory frameworks in these regions may not align with the review's objectives. Further refining the scope, the literature was published within the last ten years to capture recent advancements and emerging trends in the rapidly evolving field of AI technology. Keywords such as "artificial intelligence," "radiology," "ethics," "liability," and "implementation" guided the selection of studies directly addressing AI applications in radiological practices, workflow optimization, patient outcomes, and associated legal and ethical considerations. By concentrating 38 on these delimitations and the most recent literature, this review aims to provide a precise, pertinent, and up-to-date analysis of AI's role in radiology, highlighting its advantages and challenges in advanced healthcare systems. Ethical Assurances Ensuring ethical integrity in a literature review requires critically evaluating several key components to uphold the highest ethical standards. This includes identifying potential biases and conflicts of interest in the authors' publications, ensuring proper authorship and acknowledgment, and assessing data integrity by checking for any signs of falsification or fabrication. Additionally, publication ethics were considered by identifying redundant or duplicated articles. These elements were carefully reviewed and structured to support the analysis in an unbiased manner. The articles selected focused on examining the current state of Artificial Intelligence in well-developed countries, with a focus on enhancing patient care and satisfaction through AI integration while also safeguarding jobs in the healthcare sector. Since this study is a literature review, certain ethical considerations were not applicable, such as obtaining informed consent, securing ethical approval from an Institutional Review Board (IRB), ensuring confidentiality and privacy for human participants, protecting the rights and dignity of vulnerable populations, and following ethical guidelines for the humane treatment of animals in any research involving animal subjects. Summary The purpose of this thesis is to examine the challenges and opportunities associated with implementing AI in the US healthcare system, with a specific focus on ethical considerations and 39 legislative constraints. The aim is to develop a comprehensive understanding of how AI can be strategically integrated into radiology to improve patient care. To address concerns about potential job displacement among radiologists, this thesis explores ways AI can be used as a collaborative tool to enhance healthcare professionals’ capabilities rather than replace them. Through an in-depth analysis, this research provides insights into navigating the intricate balance between technological advancement, ethical imperatives, and the preservation of critical human roles in the healthcare landscape. 40 Chapter 4: Findings Artificial intelligence (AI) presents significant opportunities in radiology, enhancing diagnostic accuracy, workflow efficiency, and improving patient care. In the United States, AI integration within imaging modalities, such as cardiac imaging and CT for vascular and stroke assessments, is becoming increasingly common. This trend is also evident globally, particularly in Europe and other countries, where the utilization of AI is becoming more frequent. Although AI has shown promise in improving patient outcomes, AI faces several barriers. Concerns of legal liability and ethical implications pose significant obstacles to the integration of AI in the United States. In the future, AI will play a pivotal role in automating repetitive tasks and streamlining workflows, allowing radiologists to focus on more complex cases and limiting the volume of cases needing to be analyzed. Results The introduction of AI into the radiology field is an opportunity to enhance the overall healthcare workflow of radiologic imaging and how it is processed. One of the major advantages is the potential to improve diagnostic accuracy and efficiency from things such as patient scheduling, to the radiologist processing and interpretation of images (Wang & Summers, 2012). Within the accuracy of AI, deep learning algorithms are created for the technical aspect of imaging as well as the analysis of a large amount of images. This assists the radiologist in detecting abnormalities and can create a more confident diagnosis (Chartrand, 2017). This alone will improve the patient outcome and improve workflow efficiency by reducing reading times and minimizing errors. AI-driven workflow management tools can enhance efficiency by assisting with appointment scheduling and sending patient reminders which will lead to a better 41 patient experience and streamline workflow (van Leeuwen et al., 2021a). Our findings suggest that incorporating AI may enhance the capabilities of radiologists in their current role, moving from traditional pattern recognition to data management and interpretation. This implies a need to revisit theories of professional practice in healthcare, emphasizing continual learning and adaptation to emerging technologies such as AI. Incorporating AI into radiology workflows also underscores the importance of understanding how technology shapes professional roles and responsibilities within the medical field. AI-driven tools present opportunities to improve diagnostic accuracy, streamline workflows, and enhance patient care outcomes. AI can optimize imaging protocols, minimize errors, and boost productivity, thereby enhancing overall practice efficiency (van Leeuwen et al., 2021a). Evaluation of Findings With all the promise that AI brings to radiology, there are some challenges and considerations that will need to be addressed before there is a widespread use of AI. Some of the challenges are the ethical matters, liability issues, and legal aspects of AI in the healthcare system. Ethical considerations with AI deal with patient consent and data privacy and security. Patient data will need to be the main concern for the ethical standard and patient trust within the system. Another challenge with AI is the lack of human-to-human interaction which takes away from the empathy and compassion shown within healthcare settings (Furhad, 2021). The liability issues come about in the accountability in the case-related errors and even malpractice. Ethically, each facility would have to determine roles of responsibility and develop a standard at which legal responsibility can be referenced in situations (Maliha, 2021). The AI systems will need to have guidelines and regulations to ensure that patient safety and data security are at the 42 forefront. There must be a balance in the improvement AI makes in the accuracy and efficiency and the safety of patient’s diagnosis and prognosis and the security of data in the system. Looking into the future of AI in radiology, there is great promise in the innovation it will bring to healthcare and the processing of imaging. There are still challenges that will continue to be addressed and corrected. AI research and development should focus on the advancement of and structuring of algorithms and building a reputation for a secure and reliable data-sharing system (Dewey, 2020). Another future requirement of AI manufacturers will be to have training and educational lectures to help solidify the development of AI and its use in healthcare. International knowledge sharing will help accelerate AI programs and algorithms in other countries. Germany has been the frontrunner in utilizing AI programs in medical imaging and serves as a model for other countries to follow and learn from their implemented guidelines (MTR Consult, 2019). Once countries resolve the legal and ethical issues and begin to adapt to the changes and challenges of AI systems, these methods will become valuable tools used in correlation with healthcare systems and radiologists alike. Summary In conclusion, AI has the potential to transform the imaging practice in healthcare with the quality, efficiency, and accuracy of the way radiology is currently perceived in the delivery of patient care. Through the utilization of machine learning and deep learning algorithms, radiologists can now swiftly and accurately analyze medical images, assisting in the detection of abnormalities and facilitating faster diagnoses with more efficient care being given to the patient. This growing application will be accompanied by challenges concerning ethical and legal considerations, especially regarding regulatory frameworks, evidence-based assessments, and 43 liability concerns, particularly in the United States. Ethical considerations surrounding AI in healthcare, such as informed consent and patient privacy, must be addressed. Collaboration between clinicians, technology developers, regulatory bodies, and patients is crucial to navigate these challenges effectively and maximize the benefits of AI in radiology. 44 Chapter 5: Implications, Recommendations, and Conclusions Implications The literature review presented here illuminates the transformative potential of artificial intelligence (AI) to catalyze revolutionary shifts in radiology practices, fundamentally altering the landscape of healthcare delivery in the United States. AI-driven technologies have emerged as formidable assets, augmenting diagnostic accuracy, refining workflow efficiency, and harboring the capacity to enhance patient outcomes across a spectrum of imaging modalities substantially. As AI continues to evolve, its integration promises further advancements in diagnostic capabilities and workflow efficiency, contributing to improving patient care. Demonstrating remarkable efficacy across various nations, including the United States and analogous socio-economic counterparts, AI applications in radiology have exhibited proficiency in pathology detection, imaging protocol optimization, and operational streamlining. From predicting patient appointment attendance in the Netherlands (van Leeuwen et al., 2023), to facilitating the early detection of critical cases like COVID-19 pneumonia in Italy (Mushtaq et al., 2021), AI stands as an indispensable ally, bolstering the capabilities of radiologists and strengthening clinical decision-making processes. Indeed, numerous European nations and China have embraced AI within radiology, showcasing tangible enhancements in detection rates, workflow optimization, and patient outcomes (Mudgal et al., 2020). AI has already permeated various imaging modalities in the United States, such as CT and X-ray procedures, yielding palpable dividends in workflow efficiency and diagnostic precision (Cowen, 2023). Nevertheless, substantial legal liabilities and ethical quandaries persist, impeding widespread adoption and necessitating concerted efforts to address these challenges as AI becomes 45 increasingly ubiquitous in American healthcare practices, mirroring its trajectory in comparable nations. Although the immense potential of AI in radiology is evident, the United States has been cautious in embracing it, lagging behind leading countries like Germany, primarily because of complex legal and ethical challenges. While overseas deployments have showcased considerable potential in augmenting diagnostic accuracy and workflow efficiency, uncertainties loom large concerning the ethical implications, legal ramifications, and enduring impacts on the fabric of the US healthcare system. Therefore, conducting thorough investigations and fostering collaborative efforts among policymakers, healthcare stakeholders, technology developers, and regulatory entities are essential and imperative to navigate these complexities astutely. Ethical considerations, encompassing issues such as patient consent, data privacy, and the preservation of interpersonal interactions, represent formidable obstacles necessitating meticulous deliberation prior to the widespread integration of AI into American radiology practices (Furhad, 2021). Recommendations A pivotal impediment hindering the seamless assimilation of AI into radiological workflows within the United States is the intricate labyrinth of liability concerns. Ambiguities surrounding accountability in diagnostic errors or malpractice involving AI systems erect significant barriers to adoption (Maliha, 2021). The developing state of legal frameworks and regulatory protocols governing AI in healthcare constrains its role and expansion within the US healthcare infrastructure. Formulating clear guidelines and regulations is paramount to ensure patient safety and data integrity, striking a delicate equilibrium between AI's strides in precision 46 and efficiency and the imperative of safeguarding patient information. Achieving this balance mandates collaborative endeavors among stakeholders to develop ethical benchmarks and legal frameworks that accord primacy to patient welfare while harnessing AI's transformative potential in radiology. International collaboration and knowledge exchange, epitomized by trailblazers like Germany, furnish invaluable insights and models for other nations, including the United States, to emulate (Doshi, 2023). This underscores each stakeholder's indispensable role in leveraging technological advancements to enhance patient care. Looking ahead, the horizon of AI in radiology indicates further advancements in diagnostic capabilities, workflow streamlining, and patient outcomes. As AI continues its evolutionary path, confronting ethical dilemmas, validation concerns, data sharing issues, and educational needs is pivotal for seamless integration into clinical practice (Dewey, 2020). Conclusion Integrating AI into radiology practices harbors immense potential for transforming healthcare delivery in the United States. Despite significant strides witnessed globally and the tangible benefits evidenced across various imaging modalities, enduring legal and ethical concerns impede widespread adoption. Addressing challenges regarding liability issues, ethical considerations, and regulatory frameworks is imperative to unlock AI's full potential in enhancing diagnostic accuracy, workflow efficiency, and patient outcomes. International collaboration, underscored by exemplars like Germany, offers invaluable guidance for navigating these challenges. Considering the future, sustained efforts to confront ethical, legal, and technical considerations will be indispensable for the responsible and effective assimilation of AI into 47 radiology practices, ultimately fostering innovation in healthcare delivery and advancing patient care. 48 References Akpakwu, E., & Bernaert, A. (2018, May 31). Four ways AI can make healthcare more efficient and affordable. World Economic Forum. https://www.weforum.org/agenda/2018/05/four-ways-ai-is-bringing-down-the-cost-ofhealthcare/ Barat, M., Boeken, T., Duron, L., Feydy, A., Feydy, J., Lecler, A., & Soyer, P. (2023). Artificial Intelligence in Diagnostic and interventional radiology: Where are we now? Diagnostic and Interventional Imaging, 104(1), 1–5. https://doi.org/10.1016/j.diii.2022.11.004 Becker, C. D., Kotter, E., Fournier, L., & Martí-Bonmatí, L. (2022). Current practical experience with Artificial Intelligence in Clinical Radiology: A survey of the European Society of Radiology. Insights into Imaging, 13(1). https://doi.org/10.1186/s13244-022-01247-y Chartrand, G., Cheng, P. M., Vorontsov, E., Drozdzal, M., Turcotte, S., Pal, C. J., Kadoury, S., & Tang, A. (2017). Deep learning: A Primer for Radiologists. RadioGraphics, 37(7), 2113–2131. https://doi.org/10.1148/rg.2017170077 Chatterjee, A., Somayaji, N. R., & Kabakis, I. M. (2019). Abstract WMP16: Artificial Intelligence detection of cerebrovascular large vessel occlusion - nine month, 650 patient evaluation of the Diagnostic Accuracy and Performance of the Viz.ai LVO algorithm. Stroke, 50(Suppl_1). https://doi.org/10.1161/str.50.suppl_1.wmp16 Chong, L. R., Tsai, K. T., Lee, L. L., Foo, S. G., & Chang, P. C. (2020). Artificial Intelligence Predictive Analytics in the management of outpatient MRI appointment no-shows. American Journal of Roentgenology, 215(5), 1155–1162. 49 https://doi.org/10.2214/ajr.19.22594 Dewey, M., Dreyer, K., Langlotz, C., Niessen, W., Prainsack, B., Recht, M. P., & Smith, J. J. (2020). Integrating Artificial Intelligence into the clinical practice of radiology: Challenges and recommendations. European Radiology, 30(6), 3576–3584. https://doi.org/10.1007/s00330-020-06672-5 Farhud, D. D., & Zokaei, S. (2021). Ethical issues of Artificial Intelligence in medicine and Healthcare. Iranian Journal of Public Health, 50(11), i–v. https://doi.org/10.18502/ijph.v50i11.7600 GE Healthcare. (2023, February 28). AI in X-ray: Workflow solutions to improve care delivery | GE Healthcare (United States). General Electric Company. https://www.gehealthcare.com/insights/article/ai-in-xray-workflow-solutions-to-improvecare-delivery GE Healthcare. (n.d.). Cardiq Suite | GE Healthcare (United States). General Electric Company. https://www.gehealthcare.com/products/advanced-visualization/all-applications/cardiqsuite Hickman, S. E., Payne, N. R., Black, R. T., Huang, Y., Priest, A. N., Hudson, S., Kasmai, B., Juette, A., Nanaa, M., Aniq, M. I., Sienko, A., & Gilbert, F. J. (2023). Mammography Breast Cancer Screening Triage Using Deep Learning: A UK Retrospective Study. Radiology, 309(2), e231173. https://doi.org/10.1148/radiol.231173 Hosny, A., Parmar, C., Quackenbush, J., Schwartz, L. H., & Aerts, H. J. (2018). Artificial Intelligence in radiology. Nature Reviews Cancer, 18(8), 500–510. 50 https://doi.org/10.1038/s41568-018-0016-5 Karamchandani, R. R., Helms, A. M., Satyanarayana, S., Yang, H., Clemente, J. D., Defilipp, G., Strong, D., Rhoten, J. B., & Asimos, A. W. (2023). Automated detection of intracranial large vessel occlusions using Viz.ai software: Experience in a large, Integrated Stroke Network. Brain and Behavior, 13(1). https://doi.org/10.1002/brb3.2808 Kheiron Medical. (2021). MIA: Revolutionising Mammography. Kheiron Medical Technologies. https://www.kheironmed.com/mammography/ Luca Pio Stoppino, Stefano Piscone, Saccone, S., Ciccarelli, S. A., Marinelli, L., Milillo, P., Gallo, C., Luca Macarini, & Vinci, R. (2024). Vertebral and Femoral Bone Mineral Density (BMD) Assessment with Dual-Energy CT versus DXA Scan in Postmenopausal Females. Journal of Imaging, 10(5), 104–104. https://doi.org/10.3390/jimaging10050104 Maliha, G., Gerke, S., Cohen, I. G., & Parikh, R. B. (2021). Artificial Intelligence and liability in Medicine: Balancing Safety and Innovation. The Milbank Quarterly, 99(3), 629–647. https://doi.org/10.1111/1468-0009.12504 Meyer, B. C., Meyer, D. M., St. Germain, E., Pham, N. Alwood, B. T., Van Orden, K., Wood, C., Bavarsad, R., Agrawal, K., Modir, R., Hemmen, T., Pannell, S., Longhurst, C.A., & Khalessi, A. A. (2024, September 26). (STRokE DOC-AI): Leveraging AI Tools to Optimize Both Hub and Spoke in a Telestroke Network. NEJM AI. https://ai.nejm.org/doi/full/10.1056/AI-S2400848 MTR Consult. (2019, October 14). New artificial intelligence imaging program launched in 51 Germany. Med Tech Reimbursement Consulting. https://mtrconsult.com/news/newartificial-intelligence-imaging-program-launched-germany Mudgal, K. S., & Das, N. (2020). The ethical adoption of Artificial Intelligence in Radiology. BJR|Open, 2(1), 20190020. https://doi.org/10.1259/bjro.20190020 Mushtaq, J., Pennella, R., Lavalle, S., Colarieti, A., Steidler, S., Martinenghi, C. M., Palumbo, D., Esposito, A., Rovere-Querini, P., Tresoldi, M., Landoni, G., Ciceri, F., Zangrillo, A., & De Cobelli, F. (2020). Initial chest radiographs and Artificial Intelligence (AI) predict clinical outcomes in COVID-19 patients: Analysis of 697 Italian patients. European Radiology, 31(3), 1770–1779. https://doi.org/10.1007/s00330-020-07269-8 Pacchiano, F., Tortora, M., Criscuolo, S., Jaber, K., Acierno, P., De Simone, M., Tortora, F., Briganti, F., & Caranci, F. (2024). Artificial intelligence applied in acute ischemic stroke: from child to elderly. La Radiologia Medica, 129(1), 83–92. https://doi.org/10.1007/s11547-023-01735-1 Patel, C. (2023, March 14). Germany artificial intelligence (AI) in medical imaging market report 2022 to 2030. Insights10. https://www.insights10.com/report/germany-artificialintelligence-ai-in-medical-imaging-market-analysis/ Real-world performance of Behold.ai’s red dot® chest x-ray solution. (2022, August 5). Behold.ai. https://beholdai.wordpress.com/2022/08/05/somerset_case_study/ Reuter, E. (2022, November 7). 5 takeaways from the FDA’s list of AI-Enabled medical devices. MedTech Dive. https://www.medtechdive.com/news/FDA-AI-ML-medical-devices-5takeaways/635908/ 52 Siemens AG (Aktiengesellschaft). (2020). AI COVID-19. Siemens Healthineers. https://www.siemens-healthineers.com/medical-imaging/digital-transformation-ofradiology/ai-covid-19-algorithm Tam, M. D. B. S., Dyer, T., Dissez, G., Morgan, T. N., Hughes, M., Illes, J., Rasalingham, R., & Rasalingham, S. (2021). Augmenting lung cancer diagnosis on chest radiographs: Positioning Artificial Intelligence to improve radiologist performance. Clinical Radiology, 76(8), 607–614. https://doi.org/10.1016/j.crad.2021.03.021 Tsai, I.-T., Tsai, M.-Y., Wu, M.-T., & Chen, C. K.-H. (2015). Development of an automated bone mineral density software application: Facilitation Radiologic Reporting and improvement of accuracy. Journal of Digital Imaging, 29(3), 380–387. https://doi.org/10.1007/s10278-015-9848-7 van Leeuwen, K. G., de Rooij, M., Schalekamp, S., van Ginneken, B., & Rutten, M. J. (2021a). Artificial Intelligence in Radiology: 100 commercially available products and their scientific evidence. European Radiology, 31(6), 3797–3804. https://doi.org/https://doi.org/10.1007/s00330-021-07892-z van Leeuwen, K. G., de Rooij, M., Schalekamp, S., van Ginneken, B., & Rutten, M. J. (2021b). How does Artificial Intelligence in radiology improve efficiency and health outcomes? Pediatric Radiology, 52(11), 2087–2093. https://doi.org/10.1007/s00247-021-05114-8 van Leeuwen, K. G., de Rooij, M., Schalekamp, S., van Ginneken, B., & Rutten, M. J. (2023). Clinical use of Artificial Intelligence products for radiology in the Netherlands between 2020 and 2022. European Radiology, 34(1), 348–354. https://doi.org/10.1007/s00330- 53 023-09991-5 Wang, S., & Summers, R. M. (2012). Machine Learning and Radiology. Medical Image Analysis, 16(5), 933–951. https://doi.org/10.1016/j.media.2012.02.005 54 Appendix A Screening Flow Chart per 2020 PRISMA Guidelines 55 Appendix B Highlighted Mammographic Abnormalities Detected by MIA This image illustrates Mammography Intelligent Assessment (MIA) highlighting suspicious areas on mammograms indicated by yellow and pink arrows, supporting radiologists in breast cancer detection. 56 Appendix C AI Integration of Red Dot®CXR in NHS Lung Cancer Pathway The red dot®CXR platform integrates AI into the NHS lung cancer pathway, rapidly classifying chest X-rays and prioritizing suspicious cases, reducing radiologist workload by 15% and enabling diagnosis in as little as 1 hour and 40 minutes with expedited workflows. 57 Appendix D Red Dot®CXR Accelerating Diagnostics with Heatmap Insights This image highlights the red dot®CXR platform's heatmap feature, which aids in identifying abnormalities on chest X-rays and streamlines the diagnostic process, reducing the timeline from 7 days to 3 days to enhance radiology workflow efficiency. 58 Appendix E Lung Analysis with Siemens' CT Pneumonia Analysis2 This image depicts Siemens' CT Pneumonia Analysis2, an advanced AI-driven tool for analyzing chest CT scans to quantify lung abnormalities associated with COVID-19. By providing comprehensive metrics such as opacity scores and lobe-specific involvement, the tool facilitates precise evaluation of disease severity and progression, optimizing the accuracy of clinical decision-making. 59 Appendix F BMD Reporting with LiberaBMD Using Hologic DXA Data This image demonstrates the LiberaBMD automated reporting system, which utilizes raw data from a Hologic Delphi DXA scanner to generate bone mineral density (BMD) reports, including T-scores, Z-scores, and fracture risk assessments, providing a streamlined and detailed evaluation of osteoporosis and overall bone health. 60 Appendix G GE CardIQ Suite Utilizing AI for CAD Detection This image illustrates the GE CardIQ Suite utilizing advanced artificial intelligence to enhance the detection and assessment of coronary artery disease (CAD) by integrating CT imaging with detailed anatomical visualization, plaque burden analysis, and colorcoded perfusion mapping to highlight pathological findings. 61 Appendix H Viz.ai LVO Detection and Workflow Optimization This image illustrates the Viz.ai platform's use of advanced machine learning to automate large vessel occlusion (LVO) detection and optimize workflows within the extended stroke radiographic assessment (ESRA) window. By analyzing CT and CTA for LVO and core-penumbra mismatch while facilitating centralized communication, it reduces transfer times by 50%, accelerating decision-making and enhancing thrombectomy management in the critical 0–24 hour window.