Disruption or Evolution? Will AI Reshape Medical Imaging?
By Yosefa Pessin, DHSc, RDMS, RDCS, RVT | Associate Professor, Diagnostic Medical Imaging, SUNY Downstate Medical Center
Radiologic imaging is dependent on scientific advancements, high-tech equipment, technologists’ skills, patient cooperation, and radiologists’ expertise. As equipment, imaging techniques, and medicine evolve, standards of care and protocols change over time. On average, it takes about 10 years for radiology to adopt and integrate modern technologies and applications, for better or for worse.
Remember when imaging was analog, images were acquired using film, cassettes and chemical processors? Digital imaging, 3D imaging, laser printing, and finally DICOM imaging and PACS came next. Artificial Intelligence (AI) is the newest frontier. Although AI has been in existence for quite some time, the creation and adoption of AI tools for healthcare have developed at a slow pace. Radiology has had some computer-assisted detection tools for decades, but they were extremely specific. Only recently have we seen a resurgence in the development of new radiologic AI tools, which started to gain momentum during the COVID pandemic. Artificial intelligence has the potential to be a game changer in the healthcare arena. The most basic applications already available and being adopted include AI assistive technologies to provide support with patient scheduling, billing, and ambient AI for clinical notes. These types of programs are meant to reduce administrative burden, and studies are examining whether they can reduce burnout and/or boost productivity. But what about the design, adoption, and integration of radiologic AI tools? Are medical centers adopting tools to screen for specific pathologies, serve as computer-assisted detection, or provide risk analysis for disease? Are engineers working with hospital administration and healthcare professionals to develop these tools from the ground up?
Systems should provide augmented intelligence, and should not replace clinicians’ decisions, as AI should serve in assistive and advisory roles, and not as a substitute for the physician.
The American Medical Association (AMA) conducted a survey in 2024 that investigated physicians’ impressions of the integration of AI tools into healthcare delivery. Sixty eight percent reported finding value, and 66% of respondents in that study were already using some type of AI tool (AMA, 2025). The AMA also established policy regarding AI development, deployment and use in healthcare, addressing transparency, ethical, responsible, and equitable use. Their policy addresses government oversight, when and what to disclose to advance AI transparency, generative AI policies, physician liability for use of AI-enabled technologies, concerns around data privacy and payor use of AI systems.
The conditions that need to be met to ensure that healthcare professionals’ collaboration with AI tools improves patient care without the expense of autonomy and trust are many. The creation of tools must be based on randomized pragmatic trials that, within their design, have low to no bias, incorporate both the needs of patients and clinicians, account for unintended consequences, and provide an efficient method of deployment with effective usable results. Steps should be taken to remove the fear of ‘black box’ model issues, to promote greater trust by physicians and patients alike. Clinical evidence must support the adoption of such tools.
There is great potential for developing AI tools for imaging, as well as predictors of future disease using algorithmic tools, but these are in their infancy, and one must consider how the training data was acquired; does the data represent the average population or only those who chose to seek medical care if the tool was designed in a retrospective manner. There is a push to begin using AI to help facilitate quicker time to diagnoses for patients, with new tools screening for disease, and serving in an ‘advisor’ capacity for physicians and other health care professionals. The goal is to reduce the burden on healthcare providers, while providing decision support to streamline time to diagnosis and accurate, efficient patient care. Are radiology departments on board? Do radiologists and technologists want to be involved in the development of AI tools? Until now, tools have been developed in silo’s for specific organs or pathologies, or integrated into equipment to facilitate quicker scanning times.
Knowing how an AI tool will impact workflow when integrated or adopted should be discussed at the design phase. It should not result in extra effort and time, but should fit seamlessly in the daily care workflow. Many times, this is part of the obstacle to adopting AI tools in healthcare environments. Providers should be part of a feedback loop to report back about the success or challenges to the tool’s use and implementation. Concerns for accuracy and alert fatigue should be addressed as well. Other barriers that must be considered include integration into existing systems such as electronic health records, insurance, funding, liability, collaboration between various health systems, resistance to change and confidence in AI tools (World Economic Forum, 2025). What tools are currently being evaluated or developed? Who is training these AI algorithms or models? There are ethical and legal considerations in using patient data to train these AI algorithms; do the patients agree to the collection and storing of their data for this use? Is the data encrypted and protected from exposure to personally identifiable health information, is it safe from cybercriminals? These are all valid concerns. Have you been asked to help train an AI agent by quantifying details in imaging?
Systems should provide augmented intelligence, and should not replace clinicians’ decisions, as AI should serve in assistive and advisory roles, and not as a substitute for the physician (AMA, 2025). The potential for AI in healthcare is not yet fully realized and is behind other sectors of industry in development, adoption, and integration (World Economic Forum, 2025). Change is on the horizon. Are you going to get involved or watch on the sidelines? AI tools will only be as accurate and reliable as the data they are trained on. It is time to consider not if AI will impact radiologic imaging, but how. As an imaging professional, stop and consider, what clinical challenges you want support with? Think about how an AI tool can help improve patient care. Then decide how you may want to approach the challenge.
References:
American Medical Association. (2024). Augmented Intelligence Development, Deployment, and Use in Health Care. Retrieved from https://www.ama-assn.org/system/files/ama-ai-principles.pdf
American Medical Association. Augmented Intelligence in Medicine, 2025. Retrieved from https://www.ama-assn.org/practice-management/digital-health/augmented-intelligence-medicine
World Economic Forum. The Future of AI-Enabled Health: Leading the Way. White Paper, 2025. Retrieved from https://reports.weforum.org/docs/WEF_The_Future_of_AI_Enabled_Health_2025.pdf