photo credit: Mart Production / Pexels

Key Takeaways

• Artificial intelligence is increasingly being integrated into radiology to support image analysis, workflow efficiency, and diagnostic processes.
• Radiology departments are developing physician-led governance systems to evaluate and monitor AI tools in clinical environments.
• AI can help radiologists manage growing case volumes by automating lower-value tasks and improving triaging capabilities.
• The rapid expansion of FDA-approved radiology AI products has created challenges related to quality assessment and implementation.
• Research suggests AI may improve performance in some situations while potentially reducing diagnostic accuracy in others, highlighting the need for careful personalization and oversight.

Dr. Damon Deteso is a diagnostic radiologist at Millennium Medical Imaging in Saratoga Springs, New York, where he works with CT, MRI, ultrasound, X-ray, and nuclear medicine technologies. Damon Deteso has also contributed to the growing intersection of radiology and artificial intelligence through his consulting work with Imagen Technologies, which focused on AI interpretations of X-ray images. His background includes a physics degree from Holy Cross University, medical training at the University of Massachusetts Medical School, and cross-sectional imaging training at the University of California, San Francisco, and San Francisco General Hospital.

In addition to his clinical practice, he participates in the American College of Radiology and attends industry conferences to stay informed about emerging technologies and evolving applications of AI in medical imaging.

The Impact of AI on Radiology

In recent years, artificial intelligence (AI) has impacted virtually every industry in the United States, including the medical field. According to the American Medical Association, more than 80 percent of physicians use AI in a professional capacity, more than double the number of doctors who used it in 2023. While physicians generally use AI to summarize medical research and clinical-care documentation, Harvard Medical School has promoted the idea that AI can support far more radical transformations within the US health care system. Radiologists, for example, have started using AI in several ways.

Johns Hopkins Medicine began exploring the potential of AI in radiology several years ago. The radiology faculty has collaborated with Johns Hopkins operational managers to form the Radiology Artificial Intelligence Development (RAID) Subcommittee, part of the Department of Radiology’s Value Added Analysis Steering Committee. Members of the RAID team closely follow numerous clinical and research initiatives involving AI in radiology, with a goal of developing an effective, physician-led governance structure to assess, prioritize, implement, and continuously monitor AI usage in radiology practices.

A common use of AI in the field of radiology involves algorithms to evaluate huge sets of data and complex medical images. Proper execution may allow radiologists to enhance triaging processes, identify abnormal images with greater accuracy, and generally improve diagnostic competence. However, experts remain wary of the highly variable quality of AI services, preventing the widespread adoption of algorithms and related services in clinical environments. Andrew Menard, the executive director of radiology strategy and innovation at Johns Hopkins and a member of the RAID subcommittee, pointed out that there are upwards of 400 radiology AI products on the market with Food and Drug Administration (FDA) approval; the FDA approves more products every month, making it difficult for medical professionals to keep up.

Despite the challenge, Menard said that AI can function as a valuable tool for radiologists facing a growing volume of cases and increased clinical demands. Well-developed AI products allow medical professionals to automate lower-value processes and focus their skills on more pressing priorities.

At the Radiological Society of North America’s annual meeting in 2024, leaders in the field discussed the unique opportunities and challenges presented by AI, with a focus on how radiologists and program leaders can harness AI technology to improve patient outcomes. RSNA President Curtis P. Langlotz, MD, PhD, pointed out how the field of radiology is intrinsically tied to technological advances, initially by embracing X-ray and magnetic resonance imaging (MRI) technology, and more recently, by interfacing with AI.

That said, Dr. Langlotz was quick to point out the limitations of AI. “Anyone who works with AI knows that machine intelligence is different, not better than human intelligence.” He went on to propose several strategies for optimizing AI integration, including improved data accessibility and diversity, universal electronic image exchange, and allowing patients to contribute protected data to AI research and training programs.

Despite the relative optimism surrounding the use of AI in radiology, several studies show that AI may, in fact, hinder physician performance. Research published in Nature Medicine found that, while some radiologists benefit from AI assistance, AI technology can also prove detrimental to both a radiologist’s performance and the accuracy of imaging results. Based on this, facilities considering AI must also develop plans for personalizing assistive AI systems to each radiologist.

FAQs

How is AI currently used in radiology?

AI is commonly used in radiology to analyze medical images, assist with triaging cases, automate repetitive tasks, and support diagnostic decision-making. These tools are designed to help radiologists work more efficiently while managing increasing clinical demands.

Why are hospitals creating AI governance groups for radiology?

Hospitals and radiology departments are establishing governance structures to evaluate, prioritize, implement, and monitor AI technologies responsibly. These groups help ensure that AI systems are safe, effective, and aligned with clinical standards.

What challenges exist with AI adoption in radiology?

One major challenge is the rapidly growing number of FDA-approved AI products, which makes quality assessment and product evaluation difficult for healthcare providers. Variability in performance across AI systems also remains a concern.

Can AI replace radiologists?

Most experts view AI as a support tool rather than a replacement for radiologists. Human expertise remains essential for interpreting results, making clinical judgments, and managing complex patient care decisions.

Does AI always improve radiology performance?

No. Some studies suggest that AI can improve efficiency and diagnostic support in certain situations, while in other cases it may negatively affect physician performance or imaging accuracy. Proper oversight and customization are important for successful implementation.

About Damon Deteso

Dr. Damon Deteso is a diagnostic radiologist with Millennium Medical Imaging in Saratoga Springs, New York. Since 2004, he has worked across multiple hospitals providing imaging services that include CT, MRI, ultrasound, X-ray, and nuclear medicine procedures. He has also supported artificial intelligence initiatives through consulting work involving AI-assisted X-ray interpretation. Dr. Deteso holds a physics degree from Holy Cross University and earned his medical degree from the University of Massachusetts Medical School.