The rapid advancement of artificial intelligence (AI) technologies has revolutionized numerous fields, with healthcare being one of the most prominent areas of impact. In radiology, AI—particularly machine learning (ML) and deep learning (DL)—has introduced transformative capabilities that assist in diagnostic accuracy, efficiency, and predictive analysis (1,2). AI models, fueled by vast amounts of medical imaging data, are increasingly used to identify patterns that may be imperceptible to the human eye, enhancing radiologists’ ability to detect and interpret complex abnormalities in imaging modalities such as X-rays, computed tomography (CT), magnetic resonance imaging (MRI), and ultrasounds, ultimately leading to improved patient outcomes (3,4). 

The integration of AI, specifically ML and DL, into medical imaging has gained significant momentum over the past decade. Machine learning algorithms, including supervised, unsupervised, and reinforcement learning techniques, analyze data to develop predictive models, while deep learning, a subset of ML, employs neural networks to perform tasks such as image classification, segmentation, and anomaly detection (1,5,6). The increasing availability of high-quality imaging datasets, coupled with advancements in computational power, has propelled these technologies into clinical practice, where they are now used to augment decision-making in diagnosing diseases like cancer, cardiovascular conditions, and neurological disorders (2,7).

A bibliographic analysis of AI applications in radiology highlights the trends and growth within this field, offering insights into research patterns, publication volume, and thematic concentrations. Examining these trends enables a deeper understanding of how AI tools have evolved over time and what areas of medical imaging are benefitting most from AI integration (4,8). It also uncovers challenges that still need to be addressed, such as the need for larger datasets, improved generalizability of models, and regulatory hurdles for clinical implementation (3,9).

This study aims to conduct a comprehensive bibliographic analysis of the trends in AI applications in radiology, focusing on ML and DL technologies in medical imaging. By examining the trajectory of research publications, key focus areas, and collaborative networks, the study provides valuable insights into how AI continues to shape the future of radiology and the potential implications for clinical practice.

This bibliographic analysis assessed recent trends, applications, and challenges of artificial intelligence (AI), particularly Machine Learning (ML) and Deep Learning (DL), in radiology. A systematic search across databases such as PubMed, Scopus, and IEEE Xplore yielded 180 studies from 2019 to 2023. Using keywords like “AI in radiology,” “machine learning in radiology,” and “radiomics,” we applied inclusion and exclusion criteria, narrowing the scope to peer-reviewed studies and conference proceedings that focused on AI applications in radiology. After excluding non-relevant studies, 35 articles remained for analysis, covering topics such as AI’s impact on diagnostic accuracy, workflow efficiency, and radiology education (Table 1).

Data extraction was performed systematically, capturing essential details, including AI methodologies, application areas, and primary findings. A thematic analysis was applied, categorizing findings into five key areas: diagnostic applications, educational initiatives, radiologists' perceptions of AI, challenges in AI integration, and future recommendations. Each study was evaluated for relevance and methodological rigor, with limitations noted in the findings. While no ethical approval was required, ethical concerns such as data privacy were considered. This study acknowledges certain limitations, including the exclusion of non-English articles and diverse methodologies across studies, which may affect comparability.

Evolution of AI in Radiology

Artificial intelligence (AI) has transformed radiology over the past few decades, evolving from basic computer-aided detection (CAD) systems to sophisticated machine learning (ML) and deep learning (DL) algorithms. These advancements have led to improved diagnostic accuracy, enhanced clinical workflows, and increased automation in image analysis. This evolution is marked by several key phases, each building on the previous generation’s technological advancements and addressing the unique demands of radiology (10,11).

The foundation of AI in radiology began with the development of CAD systems in the 1980s and 1990s. CAD systems initially served as tools for assisting radiologists in detecting and diagnosing abnormalities in imaging, such as breast lesions in mammograms. These systems were rule-based, relying on manually crafted algorithms to recognize specific patterns. However, their accuracy and reliability were limited by the simplicity of their models, and they often produced high false-positive rates, requiring extensive human oversight (1,12).

In the 2000s, the introduction of machine learning algorithms marked a significant advancement in radiology. Unlike traditional CAD systems, ML algorithms could learn from data without relying on explicit programming. ML models, particularly supervised learning algorithms, used labeled data to recognize patterns and make predictions, allowing for greater accuracy and adaptability in medical imaging tasks (10,11). For example, ML algorithms could be trained on large datasets of annotated medical images to identify and classify lesions, tumors, and other abnormalities more effectively than previous systems. This shift from rule-based systems to data-driven models marked a critical turning point, as ML algorithms could continuously improve with larger and more diverse datasets (1,7).

The development of deep learning in the 2010s revolutionized AI applications in radiology. DL, a subset of ML, uses multi-layered neural networks that can automatically learn complex features from large datasets. This capacity to process and interpret vast amounts of image data led to unprecedented accuracy in tasks such as image segmentation, classification, and detection. Convolutional Neural Networks (CNNs), a type of DL architecture specifically designed for image analysis, became widely adopted in radiology due to their ability to capture intricate spatial hierarchies in images, making them highly effective for radiographic analysis (1,3). Radiomics, a related field that involves extracting high-dimensional quantitative features from medical images, further expanded the potential of DL by providing detailed insights into tumor biology, disease progression, and patient prognosis (13,14).

Radiomics, as part of the broader AI evolution in radiology, has enabled a deeper analysis of medical images by capturing quantitative features beyond what the human eye can perceive. These features can be correlated with clinical outcomes, enabling personalized treatment planning and improved prognosis (15). For instance, DL models in radiomics can identify subtle patterns associated with malignancy or disease progression in tumor characterization, leading to more accurate assessments. Scapicchio et al. (2021) describe radiomics as a field that allows for the extraction of large volumes of quantitative imaging data, often referred to as “features,” which can be analyzed and associated with genetic, clinical, and patient outcome data. Radiomics combined with DL has been pivotal in oncology, where it is used to identify subtypes of tumors, predict patient outcomes, and optimize treatment plans (1,16,17).

Today, AI applications in radiology are being integrated more seamlessly into clinical workflows, with DL models driving many of the recent advancements in image analysis. AI models are now capable of performing complex diagnostic tasks autonomously, such as detecting pulmonary nodules in chest CT scans or identifying fractures in X-rays, with high levels of accuracy. This shift toward AI-assisted diagnosis is reshaping the role of radiologists, allowing them to focus on more complex cases while reducing the time required for routine image analysis (4,18). However, this evolution has also raised concerns among radiologists regarding the potential of AI to replace human expertise in certain aspects of radiology, leading to discussions about AI’s role as a support tool rather than a replacement (2,16,17).

AI applications have also expanded beyond diagnostic support to include predictive analytics, which can help forecast patient outcomes based on imaging data and clinical history. For instance, DL models can now predict the likelihood of disease recurrence in oncology patients or the risk of developing certain conditions based on initial imaging findings. These predictive capabilities represent a significant shift in radiology, positioning AI not only as a diagnostic tool but also as an integral part of patient management and personalized care planning (19–21).

The continued evolution of AI in radiology will likely focus on integrating AI more deeply into radiologists’ workflows, enhancing its utility as a diagnostic and predictive tool, and addressing ethical concerns. Key areas of development include refining AI models to improve accuracy and reliability, expanding radiology-specific training in AI to ensure radiologists are equipped to work with these tools, and addressing regulatory challenges to ensure AI integration complies with data privacy and ethical standards(4,22). Additionally, future advancements may focus on developing hybrid AI models that combine DL algorithms with traditional radiological knowledge, allowing for more interpretable and transparent decision-making processes (23).

One critical aspect of the future evolution of AI in radiology is the need for ethical considerations around data usage, patient privacy, and the potential for AI to alter radiology’s professional landscape. Studies indicate that while AI can reduce diagnostic workload, the risk of over-reliance on AI systems, privacy concerns due to large-scale data collection, and potential bias in AI algorithms remain significant challenges (3,9). Addressing these concerns is essential to fostering a balanced relationship between AI and radiologists, with AI serving as a tool to augment rather than replace human expertise.

The evolution of AI in radiology highlights the transformative potential of technology in enhancing diagnostic precision, workflow efficiency, and patient outcomes. From early CAD systems to the current DL-powered models capable of autonomous analysis and predictive diagnostics, AI has reshaped radiology. As AI continues to evolve, balancing technological advancement with ethical, educational, and regulatory considerations will be crucial to ensuring its successful integration into radiology practice and its acceptance by medical professionals (24–26).

Current Trends in AI Applications in Radiology

In recent years, AI applications in radiology have expanded significantly, driven by advances in machine learning (ML), deep learning (DL), and radiomics. These technologies have been particularly transformative in diagnostic support, where AI algorithms enhance image analysis by improving diagnostic accuracy and efficiency. For instance, convolutional neural networks (CNNs), a subset of DL, are highly effective at tasks such as detecting pulmonary nodules on CT scans, identifying fractures on X-rays, and recognizing breast lesions on mammograms, often with accuracy comparable to human radiologists (1,3). Beyond individual diagnostics, AI is also increasingly used for population-level analyses, as it can sift through massive datasets, identifying disease patterns that may be missed in manual reviews (6,10,16).

Radiomics, another critical area in AI-driven radiology, involves extracting quantitative features from medical images that are not readily visible to the human eye. These features can be analyzed to predict disease progression and patient outcomes, thus providing insights into personalized treatment planning (1). Radiomics has been particularly beneficial in oncology, where it helps to subtype tumors and assess treatment responses. Furthermore, AI applications are now being integrated into the workflow automation of radiology departments, reducing radiologists’ workloads and enabling faster turnaround times for routine imaging tasks (4). However, the deployment of AI at scale has led to discussions about optimizing human-AI collaboration, as AI is primarily regarded as a tool to augment rather than replace radiological expertise (2,18).

Educational Initiatives and AI Training in Radiology

As AI becomes more prevalent in radiology, educational initiatives are critical to equipping radiologists and medical professionals with the necessary skills to work effectively with AI tools. Recognizing this need, some radiology programs have started to incorporate AI and ML training into their curricula. For example, AI-RADS, an artificial intelligence curriculum designed for radiology residents, is one of the pioneering models aimed at educating future radiologists on the fundamentals of AI (22). Such programs are designed to help residents understand AI algorithms, interpret AI-generated results, and apply these insights in clinical settings.

AI education for radiologists also extends to practicing professionals. Courses like "AI for Doctors" have been developed to provide continuing education on AI, focusing on how to integrate these technologies into daily clinical workflows (5). This training is essential to address the knowledge gap that exists between AI developers and healthcare practitioners, enabling radiologists to be informed users and evaluators of AI tools rather than passive recipients. Additionally, studies suggest that AI-focused education can improve radiologists’ acceptance and trust in AI, as it demystifies the technology and clarifies its role as a support tool rather than a competitor (27).

Attitudes and Perceptions of Radiologists Toward AI

Radiologists' attitudes toward AI are complex and often ambivalent, reflecting both optimism and concern. Many radiologists appreciate the potential of AI to improve diagnostic accuracy, speed up image analysis, and reduce diagnostic errors. However, concerns about AI's impact on job security are prevalent, especially among junior radiologists and radiology residents who worry that certain tasks might eventually be automated (2,4). An international survey among radiologists and radiology residents highlighted that while radiologists are generally positive about AI’s ability to support clinical tasks, they are also cautious about its limitations, such as the potential for algorithmic bias and errors that could affect patient care (21,24,28).

Different groups within radiology hold varying views on AI. For instance, medical students and surgeons often exhibit more skepticism toward AI compared to practicing radiologists, who tend to see AI as a valuable tool to enhance efficiency rather than a threat (18). In some cases, the perception of AI is also influenced by the availability of AI-specific education; radiologists who receive training on AI integration are often more receptive and optimistic about its benefits (3,6). This finding suggests that education and familiarity with AI can mitigate fears and increase confidence in its use.

Challenges and Barriers to AI Integration

Despite its promise, the integration of AI into radiology faces numerous challenges. Technical limitations are a significant barrier; AI algorithms require vast amounts of high-quality annotated data for training, and access to such data is often restricted due to privacy regulations and logistical constraints (9,28). Furthermore, AI models trained on limited datasets may not generalize well across different patient populations, leading to concerns about the reliability of AI in diverse clinical settings(29).

Ethical and privacy concerns also present challenges, especially given that AI systems often rely on sensitive patient data. Ensuring data privacy and obtaining informed consent for AI applications are crucial yet complex issues, as data usage policies and patient expectations vary widely (30). In addition, regulatory and legal barriers hinder AI implementation. Regulatory bodies have been cautious in approving AI tools due to concerns about accountability in case of diagnostic errors. Radiologists and institutions are also wary of liability issues, particularly around who holds responsibility for AI-driven diagnoses (27,31). Finally, there is resistance from within the medical community, with some radiologists reluctant to adopt AI due to concerns about accuracy, transparency, and the risk of over-reliance on these technologies.

Future Directions and Recommendations

To enhance AI’s role in radiology, several future directions and recommendations have been proposed. Improving the transparency and interpretability of AI models is essential; radiologists need to understand the reasoning behind AI outputs to trust and effectively use them. Hybrid AI models, which combine DL algorithms with traditional radiological expertise, could offer a balanced approach, harnessing AI’s efficiency while allowing human oversight for complex cases (32,33). Furthermore, interdisciplinary collaboration between AI developers, radiologists, and regulatory bodies can help address technical and regulatory challenges, ensuring that AI systems meet clinical standards and maintain patient safety (8,11,34).

Educational initiatives should continue to be a priority, with radiology programs incorporating AI training to ensure that future radiologists are equipped to work with these tools. Institutions are encouraged to provide ongoing education for practicing radiologists to keep them updated on AI advancements (7,8,22). Policy recommendations also emphasize the importance of data-sharing frameworks that protect patient privacy while enabling access to the large datasets AI models require (35). Implementing these recommendations could facilitate smoother AI integration, enabling radiology to benefit from AI’s diagnostic and workflow efficiencies without compromising ethical or professional standards.

AI has fundamentally transformed radiology, from its initial CAD systems to the current sophisticated DL models that drive diagnostic support, workflow automation, and personalized medicine. While AI applications hold immense potential to improve patient care and radiology efficiency, successful integration requires addressing both technical and ethical challenges. Current trends indicate a movement toward AI-enhanced diagnostics and predictive analytics, yet radiologists' perspectives remain varied, influenced by factors like job security, trust in AI, and access to AI-specific education. As AI continues to evolve, balancing human expertise with AI-driven insights will be essential for achieving its full potential in radiology. By investing in education, interdisciplinary collaboration, and policy development, the radiology community can ensure that AI becomes an effective, ethical, and accepted tool in clinical practice.

1. Scapicchio C, Gabelloni M, Barucci A, Cioni D, Saba L, Neri E. A deep look into radiomics. Radiol Med (Torino). 2021;126:1296–311.
2. Alsharif W, Qurashi A, Toonsi F, Alanazi A, Alhazmi F, Abdulaal O, et al. A qualitative study to explore opinions of Saudi Arabian radiologists concerning AI-based applications and their impact on the future of the radiology. BJR Open. 2022;4:null.
3. Coppola F, Faggioni L, Regge D, Giovagnoni A, Golfieri R, Bibbolino C, et al. Artificial intelligence: radiologists’ expectations and opinions gleaned from a nationwide online survey. Radiol Med (Torino). 2020;126:63–71.
4. Huisman M, Ranschaert E, Parker W, Mastrodicasa D, Kočí M, Santos DP de, et al. An international survey on AI in radiology in 1,041 radiologists and radiology residents part 1: fear of replacement, knowledge, and attitude. Eur Radiol. 2021;31:7058–66.
5. Hedderich D, Keicher M, Wiestler B, Gruber MJ, Burwinkel H, Hinterwimmer F, et al. AI for Doctors—A Course to Educate Medical Professionals in Artificial Intelligence for Medical Imaging. Healthcare. 2021;9:null.
6. Abuzaid M, Elshami W, McConnell J, Tekin H. An extensive survey of radiographers from the Middle East and India on artificial intelligence integration in radiology practice. Health Technol. 2021;11:1045–50.
7. Mehrizi MHR, Ooijen PV van, Homan M. Applications of artificial intelligence (AI) in diagnostic radiology: a technography study. Eur Radiol. 2020;31:1805–11.
8. Shelmerdine S, Rosendahl K, Arthurs O. Artificial intelligence in paediatric radiology: international survey of health care professionals’ opinions. Pediatr Radiol. 2021;52:30–41.
9. Strohm L, Hehakaya C, Ranschaert E, Boon W, Moors E. Implementation of artificial intelligence (AI) applications in radiology: hindering and facilitating factors. Eur Radiol. 2020;30:5525–32.
10. Khafaji MA, Safhi MA, Albadawi RH, Al-Amoudi SO, Shehata SS, Toonsi F. Artificial intelligence in radiology. Saudi Med J. 2020;43:53–60.
11. Chiwome L, Okojie OM, Rahman AKMJ, Javed F, Hamid P. Artificial Intelligence: Is It Armageddon for Breast Radiologists? Cureus. 2020;12:null.
12. Grassi R, Miele V, Giovagnoni A. Artificial intelligence: a challenge for third millennium radiologist. Radiol Med (Torino). 2019;124:241–2.
13. Weisberg E, Fishman E. Developing a curriculum in artificial intelligence for emergency radiology. Emerg Radiol. 2020;27:359–60.
14. Ooi SKG, Makmur A, Soon AY, Fook-Chong S, Liew C, Sia SY, et al. Attitudes toward artificial intelligence in radiology with learner needs assessment within radiology residency programmes: a national multi-programme survey. Singapore Med J. 2019;null:null.
15. Scapicchio C, Gabelloni M, Barucci A, Cioni D, Saba L, Neri E. A deep look into radiomics. Radiol Med (Torino). 2021;126:1296–311.
16. Pauwels R, Rey YCD. Attitude of brazilian dentists and dental students regarding the future role of artificial intelligence in oral radiology: a multi-center survey. Dento Maxillo Facial Radiol. 2020;null:20200461.
17. Jungmann F, Jorg T, Hahn F, Santos DP dos, Jungmann SM, Düber C, et al. Attitudes Toward Artificial Intelligence Among Radiologists, IT Specialists, and Industry. Acad Radiol. 2020;null:null.
18. Hoek J van, Huber A, Leichtle A, Härmä K, Hilt D, Tengg-Kobligk H von, et al. A survey on the future of radiology among radiologists, medical students and surgeons: Students and surgeons tend to be more skeptical about artificial intelligence and radiologists may fear that other disciplines take over. Eur J Radiol. 2019;121:108742.
19. Gong B, Nugent JP, Guest W, Parker W, Chang P, Khosa F, et al. Influence of Artificial Intelligence on Canadian Medical Students’ Preference for Radiology Specialty: ANational Survey Study. Acad Radiol. 2019;26 4:566–77.
20. Santomartino SM, Yi P. Systematic Review of Radiologist and Medical Student Attitudes on the Role and Impact of AI in Radiology. Acad Radiol. 2022;null:null.
21. Eiroa D, Antolín A, Ascanio MF del C, Ortiz VP, Escobar M, Roson N. The current state of knowledge on imaging informatics: a survey among Spanish radiologists. Insights Imaging. 2022;13:null.
22. Lindqwister AL, Hassanpour S, Lewis P, Sin JM. AI-RADS: An Artificial Intelligence Curriculum for Residents. Acad Radiol. 2020;null:null.
23. Neri E, Coppola F, Miele V, Bibbolino C, Grassi R. Artificial intelligence: Who is responsible for the diagnosis? Radiol Med (Torino). 2020;125:517–21.
24. Qurashi A, Alanazi RK, Alhazmi YM, Almohammadi A, Alsharif W, Alshamrani K. Saudi Radiology Personnel’s Perceptions of Artificial Intelligence Implementation: A Cross-Sectional Study. J Multidiscip Healthc. 2021;14:3225–31.
25. Alelyani M, Alamri S, Alqahtani M, Musa A, Almater H, Alqahtani N, et al. Radiology Community Attitude in Saudi Arabia about the Applications of Artificial Intelligence in Radiology. Healthcare. 2021;9:null.
26. Galán GC, Portero FS. Medical students’ perceptions of the impact of artificial intelligence in radiology. Radiologia. 2021;null:null.
27. Gorospe-Sarasúa L, Muñoz-Olmedo JM, Sendra-Portero F, Luis-García R de. Challenges of Radiology education in the era of artificial intelligence. Radiologia. 2022;64 1:54–9.
28. Eltorai A, Bratt A, Guo H. Thoracic Radiologists’ Versus Computer Scientists’ Perspectives on the Future of Artificial Intelligence in Radiology. J Thorac Imaging. 2019;35:255–9.
29. Yang L, Ene I, Belaghi RA, Koff D, Stein N, Santaguida P. Stakeholders’ perspectives on the future of artificial intelligence in radiology: a scoping review. Eur Radiol. 2021;32:1477–95.
30. Neri E, Coppola F, Miele V, Bibbolino C, Grassi R. Artificial intelligence: Who is responsible for the diagnosis? Radiol Med (Torino). 2020;125:517–21.
31. Waymel Q, Badr S, Demondion X, Cotten A, Jacques T. Impact of the rise of artificial intelligence in radiology: What do radiologists think? Diagn Interv Imaging. 2019;100 6:327–36.
32. Coppola F, Faggioni L, Gabelloni M, Vietro FD, Mendola V, Cattabriga A, et al. Human, All Too Human? An All-Around Appraisal of the “Artificial Intelligence Revolution” in Medical Imaging. Front Psychol. 2021;12:null.
33. Sard MLSPCCLMFCMM van KS, Codari M, Melazzini L, Morozov S, Kuijk CV van, Sconfienza L, et al. Impact of artificial intelligence on radiology: a EuroAIM survey among members of the European Society of Radiology. Insights Imaging. 2019;10:null.
34. Giansanti D, Rossi I, Monoscalco L. Lessons from the COVID-19 Pandemic on the Use of Artificial Intelligence in Digital Radiology: The Submission of a Survey to Investigate the Opinion of Insiders. Healthcare. 2021;9:null.
35. Recht M, Dewey M, Dreyer K, Langlotz C, Niessen W, Prainsack B, et al. Integrating artificial intelligence into the clinical practice of radiology: challenges and recommendations. Eur Radiol. 2020;30:3576–84