From Artificial Intelligence to Robotics: How to Navigate Technological Innovation in Radiology with the Gartner Hype Cycle

Abstract

Radiology is experiencing rapid technological transformation across multiple domains, making it challenging to distinguish innovations with genuine, lasting clinical value from those that are likely to fade after initial enthusiasm. This narrative review uses an expert-driven approach supported by targeted literature review to apply the Gartner hype cycle framework to assess the maturity and real-world evidence of major radiology advancements across three key areas: software and algorithms, advanced imaging tools and techniques, and clinical practice paradigms. We analyze ongoing trends within each of these domains and provide strategic considerations for radiologists, trainees, and leaders contemplating how to navigate this evolving landscape in the context of broader themes such as cybersecurity, equitable access, sustainability, and the implications for the specialty. Radiologists should continue to develop competencies in AI and subspecialty-relevant advanced imaging, trainees should pursue structured opportunities to engage with emerging technologies, and radiology leaders should establish robust evaluation and governance frameworks supported by scalable infrastructure, including enterprise imaging and remote work systems.

As part 6 of the Radiology Research Alliance (RRA) review series on emerging technologies in collaboration with the University of Maryland Institute for Health Computing (UM-IHC) and the Medical Intelligent Imaging (UM2ii) Center, this paper provides a critical perspective on how radiologists, trainees, and leaders can navigate the hype cycle to identify meaningful innovations and guide strategic adoption in practice.

Introduction

Throughout its history, radiology has been driven by waves of technological change, ranging from the emergence of new modalities to meet evolving clinical needs to the transformation from film-based studies to digital workflows.¹ We are in the midst of another accelerated phase of radiology innovation, with new tools rapidly proliferating in areas ranging from software to hardware. However, the sheer number and diversity of new radiology technologies can make it difficult for radiologists to separate those which deliver genuine clinical value and improve patient outcomes from those that will fade after early enthusiasm wanes. In this narrative review, we use an expert-driven approach to identify and categorize key technologies based on clinical impact, adoption trends, and prominence in professional discourse and conferences, supplemented by targeted literature review for supporting evidence. We adapt the Gartner hype cycle framework to assess several major areas of innovation in terms of their maturity and real-world evidence, with the aim of identifying which technologies are most likely to have meaningful and lasting impact.

The Gartner Hype Cycle and its Application to Radiology Innovation

The Gartner hype cycle is a subjective, high-level graphical framework that illustrates the evolution of innovations through five different phases of hype and maturity.^2,3 We present this framework as a useful heuristic for radiologists and leaders to critically evaluate emerging technologies within their specific institutional contexts.

To help illustrate the Gartner hype cycle (Figure 1A), we will define its five phases and trace a radiology example (the historical journey of the Picture Archiving and Communication Systems [PACS]) to set the stage.^4–6 This cycle closely mirrors the broader technology adoption lifecycle, categorizing users into distinct groups ranging from innovators, who eagerly experiment with emerging technologies, to early adopters, early majority, late majority, and finally laggards, who are cautious or resistant to change. Radiologists may individually identify with different adopter groups depending on personal attitudes, experience, and institutional circumstances; there is no universally correct stance. And, stances may be different depending on the type of technology. Notably, between the early adopters and the early majority lies a critical inflection point, termed the “chasm,” where many promising innovations stall due to practical integration barriers and insufficient real-world preparedness. Successfully crossing this chasm requires radiologists and departmental leaders to proactively anticipate and strategically navigate the practical “last mile” challenges inherent in technology adoption, ideally in collaboration with vendor partners. Radiologists should understand that effectively integrating new technologies into everyday clinical practice typically demands substantial workflow adjustments, targeted training, and ongoing change management. Likewise, leaders must remain vigilant about potential unforeseen hurdles - such as infrastructure limitations, user acceptance issues, and organizational cultural resistance - that can derail even the most promising innovations. Successfully navigating technological innovation in radiology demands more than mere involvement - it requires genuine commitment. Similar to the well-known parable of “The Chicken and the Pig” where in a ham-and-egg dish, the chicken is merely involved (providing eggs) while the pig is fully committed (providing bacon), radiologists and leaders must consider their degree of commitment when assessing and adopting emerging technologies. Recognizing and preparing for these implementation barriers allows radiologists and leaders to realistically evaluate timelines, mitigate risks, and balance enthusiasm for novel technologies with a thoughtful, practical approach to their adoption.

Figure 1. The Gartner hype cycle and innovation adoption lifecycle.

(A, top) The Gartner hype cycle framework five stages of technology maturity: innovation trigger, peak of inflated expectations, trough of disillusionment, slope of enlightenment, and plateau of productivity (yellow). This cycle is closely aligned with the classical technology adoption curve, which describes distinct adopter groups from innovators and early adopters, through the challenging “chasm” (red), to early majority, late majority, and laggards. Successfully crossing the chasm and reaching widespread adoption requires navigating the practical challenges of implementation and integration—often referred to as the “last mile.” (B, bottom) Applied to radiology, this framework helps to generally map the current state and adoption readiness of emerging technology innovations across three major categories: software and algorithms (blue), advanced imaging tools and techniques (purple), and clinical practice paradigms (green).

1. Innovation Trigger

   Early proof-of-concept studies spark initial curiosity. The idea of the PACS first emerged in the 1970s when academic research laboratories and some industry vendors began exploring early concepts and producing prototypes.^4,5

2. Peak of Inflated Expectations

   As external interest increases, funding surges and vendors begin producing commercial technology. The buzz surrounding PACS grew immensely as more research studies were published and demos were shown at large radiology conferences. There were ambitious plans to build filmless hospitals as early as 1982.⁶

3. Trough of Disillusionment

   As early adopters start to deploy the technology, limitations emerge. As radiology departments and practices started using commercially available PACS systems, they noticed several challenges including slow network speeds, necessity of expensive and large high-resolution monitors that deteriorated over six months, and lack of digital images for non-radiologists.^4,7 Beyond these technical challenges, institutions also faced cultural resistance to change, as radiologists accustomed to film now had to adapt to new reading environments and workflows that transformed their day-to-day practice.

4. Slope of Enlightenment

   New, improved generations of the technology are released and users better understand how to strategically employ it. As PACS users began sharing best practices as well as implementation strategies, technology improved, organizational standards emerged, and studies demonstrating clinical, operational, and financial benefits were published, PACS systems became more widely integrated.^4–6

5. Plateau of Productivity

   The technology becomes mainstream and a core part of the infrastructure. PACS systems have now of course been universally adopted and integrated into practice, and modern radiology workflows without PACS are unfathomable. New technological advancements continue to be explored.⁴

Applied to radiology, the model provides a useful framework for assessing which tools are ready for or have already achieved widespread deployment and which ones are still too early or in one of the volatile middle phases where early excitement may outweigh true impact or clinical value. In the sections below, we use the Gartner hype cycle as an illustrative guide for educational purposes and discussion, acknowledging that maturity and adoption traction can vary considerably across different radiology practice settings. Importantly, the Gartner hype cycle has well-known limitations: its categorization into stages is inherently subjective, technological innovations do not necessarily progress linearly through these stages, and the duration of each stage can vary widely based on regulatory barriers, reimbursement patterns, and differences in clinical environments including geographic and economic factors. Readers are encouraged to interpret and adapt these insights according to their own institutional contexts and personal experiences. We discuss the innovation maturity and potential future trajectory of three major areas of radiology technology trends (Figure 1B), as well as potential implications for our field and strategies for different roles.

Radiology Technology Innovations Mapped by Gartner Hype Cycle

Software and Algorithms

Overall Trends: Making Sense of the Hype Cycle

Software and algorithms represent promising computational tools designed to digest large volumes of healthcare data (Figure 2, Table 1). However, translation of computer science concepts to healthcare applications has not always been successful. For example, blockchain technology has created a stir in the financial world, but despite repeatedly being proposed as a way to safely store decentralized data and ensure patient privacy, it has not found traction in healthcare.⁸ A nascent technology to keep an eye on is quantum computing, which may more efficiently perform computations necessary for AI as well as better analyze large-scale healthcare data, however may be a decade or more away from real clinical application.⁹

Figure 2.

Applications for AI throughout the broader healthcare setting, including radiology as a collaborative diagnostic specialty. AI tools can process healthcare data across the entire patient lifecycle, such as by gathering information and boosting patient engagement pre-encounter, assisting clinicians during encounters, and helping with administrative tasks post-encounter. They can also enhance transitions between care stages while enabling advanced analytics. Success of AI tools requires thoughtful delivery science and implementation science research and application.

Table 1.

Major applications in software and algorithms organized by the Gartner hype cycle. See footnotes for definitions.

Innovation Trigger	Peak of Inflated Expectations	Trough of Disillusionment	Slope of Enlightenment	Plateau of Productivity
Quantum computing1	Automatic cardiac post-processing4	Clinical decision support tools8	Image interpretation - classification /detection11	Image optimization and reconstruction16 - Accelerated reconstruction - Denoising - Automating positioning
Synthetic/generative imaging2	Foundation models (radiology and omics)5	Cross-disciplinary care coordination9	Quantitative image analysis12	AI orchestration platforms17
Radiogenomic predictions3	Automatic image-to-report generation6	Blockchain10	Image interpretation -segmentation13	AI-assisted dictation and report completion18
	AI agents and automated multi-agent workflows7		Opportunistic screening14	Prostate post-processing19
			Simplified patient reports15	Breast CAD20

¹Quantum computing: An emerging computational paradigm leveraging quantum mechanics to perform highly complex calculations more efficiently than classical computing.

²Synthetic/generative imaging: AI techniques that create realistic medical images by synthesizing data from other modalities, sequences, or clinical inputs. For example, this can augment datasets with rare disease subtypes or create virtual noncontrast images.

³Radiogenomic predictions: AI-driven models combining imaging and genomic data for predicting clinical outcomes or treatment responses.

⁴Automatic cardiac post-processing: Computational tools automating measurement, segmentation, and analysis tasks specific to cardiac imaging.

⁵Foundation models: Large-scale AI models pre-trained on diverse, multimodal data capable of generalizing to multiple radiological tasks.

⁶Automatic image-to-report generation: AI tools automatically creating structured radiology reports directly from imaging data.

⁷AI agents and automated multi-agent workflows: Advanced AI systems capable of performing multiple sequential radiology tasks independently or semi-independently. Multi-agent workflows are those that use multiple agents for automated tasks like scientific research or multidisciplinary tumor boards.

⁸Clinical decision support tools: Software designed to enhance clinical decision-making by providing radiologists with diagnostic recommendations or alerts.

⁹Cross-disciplinary care-coordination: Systems and workflows facilitating communication and coordination across different clinical specialties.

¹⁰Blockchain: A decentralized digital ledger technology offering secure and transparent data storage and transaction verification.

¹¹Image interpretation – classification/detection: AI applications for automatically identifying and classifying abnormalities or specific findings within medical images.

¹²Quantitative image analysis: Computational methods providing numeric assessments of imaging findings for clinical interpretation and research.

¹³Image interpretation – segmentation: Automated algorithms delineating anatomical structures or pathologies within medical images.

¹⁴Opportunistic screening: Algorithms identifying incidental yet clinically significant findings from routine imaging studies not initially intended for screening.

¹⁵Simplified patient reports: Reports generated by AI tools aimed at providing clear, patient-friendly language and actionable information.

¹⁶Image optimization and reconstruction: AI-based algorithms enhancing image quality, reducing noise, and accelerating imaging acquisition and processing.

¹⁷AI platforms: Integrated software environments designed for implementing, managing, and monitoring multiple AI algorithms within clinical workflows.

¹⁸AI-assisted dictation and report completion: Software supporting radiologists by automatically generating parts of or finalizing radiology reports.

¹⁹Prostate post-processing: Dedicated tools for automated segmentation, measurement, and analysis of prostate imaging, commonly for cancer detection and management.

²⁰Breast CAD (Computer-Aided Detection): Automated systems aiding radiologists in detecting and characterizing suspicious lesions in breast imaging.

As AI technology matures and adoption climbs in healthcare settings, human factors, especially clinical-user/patient-centered design and downstream impacts on the workforce and society, are increasingly important. For example, initially, the application of narrow use-case deep learning raised concerns about radiologists being replaced by AI algorithms. Since then, there has been steady adoption of US Food and Drug Administration (FDA)-cleared AI devices focused on disease classification, detection, and segmentation,¹⁰ such as a dedicated AI tool to detect hemorrhage on a head CT, with rising interest in deploying AI platforms to orchestrate multiple tools when they are adopted. Opportunistic screening algorithms, which identify incidental findings like osteoporosis from routinely obtained clinical exams, have also grown in popularity.^11,12 The most widely deployed AI tools work behind-the-scenes to improve image quality and acquisition time, like algorithms for image optimization and reconstruction which are now widely integrated into new vendor scanner offerings.¹³ Automated post-processing has succeeded for certain well-defined tasks like prostate segmentation, but remains challenging for complex applications like cardiac MRI, which that require nuanced clinical judgment.^14,15 Thus far, the initial prediction that radiologists will be replaced has not proven true – instead, adoption of these technologies has demonstrated a synergy that requires “humans-in-the-loop,” and demand for radiologists has continued to grow over time.^16–18

However, these narrow single-task interpretive AI tools have shown mixed effects on practice metrics such as report turnaround time and overall patient outcomes.^17,19–21 More recently, advancements in machine learning models including attention-based transformers have opened the door to handling increasingly complex data, both in size and type of data. Up until recent years, AI tools were employed for only single tasks, based on a single type of information (i.e. chest CTA images for a single finding, such as a pulmonary embolism). Transformer-enabled multi-modal capabilities (i.e., handling images, text, audio, and video simultaneously) have given rise to more advanced foundation models, generative AI, and AI agents, which can now perform a wide array of tasks. These innovations can assist radiologists in tasks spanning multi-disease detection, realistic synthetic medical image data generation, and image-to-report automation.^22–25 For example, generative AI models are being used to create patient-friendly, reading-level appropriate reports or serving as an audio or textual translation service.^12,26 A prospective evaluation of generative AI-assisted radiograph reporting across a tertiary care health system demonstrated a 15.5% improvement in reporting efficiency with no difference in clinical accuracy.²⁷As foundation models and autonomous AI agents continue to mature and gain regulatory approval, they aim to fill in the gaps left by existing AI tools such as comprehensive image interpretation or multi-agent workflows to accomplish automated tasks like scientific research or tumor boards.^{22,23,28–30} Workflows combining these new technologies with human expertise promise to truly improve reporting efficiency and help balance the increasing volumes and worsening burnout in the profession.³¹

Another potential shift in the role of the radiologist may come from emerging techniques like radiogenomics and multi-omic foundation models that aim to expand personalized medicine by predicting disease prognostication and response to therapy.^28,32 The fusion of radiomics with sequencing, gene expression, and other molecular biology data may lead to radiologists working more as multidisciplinary quantitative diagnostic partners in concert with oncologists, geneticists, pathologists, and data scientists for both clinical management and translational research.

Finally, it is important to recognize that the broader AI field may be at or near the peak of inflated expectations, particularly with respect to artificial general intelligence (AGI), a form of AI which would meet or exceed human performance across all tasks. The ambitious goals of AGI development by major technology companies have generated substantial hype and bold predictions that may outpace near-term deliverables. Although AI is undoubtedly reshaping radiology and will continue to do so, maintaining realistic expectations about the timeline and capabilities of increasingly generalized AI systems will help the specialty navigate inevitable periods of disillusionment before reaching sustainable, long-term integration.

Future-thinking: Strategies for Radiologists, Trainees, and Leaders

With the growing number of available AI algorithms spanning diverse clinical applications, radiologists must make strategic decisions about which tools justify investment and how to optimize their integration into clinical practice.

Radiologists who use AI in their practice need to be cautious about becoming overly dependent on these technologies, a phenomenon known as “automation bias,” which occurs when people trust automated systems too much and overlook errors or inconsistencies.³³ It is important for radiologists to understand how to measure the accuracy and reliability of these AI tools, as responsible end-users who directly impact patient care outcomes.^34–36 They can also work closely with AI developers to test and improve these technologies by participating in validation studies (where tools are assessed for accuracy) and by providing regular feedback. This input can help shape tools through structured processes, such as vendor-led improvement plans or “Predetermined Change Control Plans,” which are currently being studied and discussed as potential ways to regulate AI products in the FDA.³⁷

Residency training should incorporate trainee education in AI, and a recent multi-society syllabus defines competencies for AI users, purchasers, clinical collaborators, and developers in radiology that can be used to help formulate a comprehensive, targeted curriculum.^38,39 Trainees do not necessarily need to learn the underlying mathematical theory, but they should have formal didactics and workshops on different radiology AI applications, learn to appraise AI performance using standard technical as well as clinical metrics, and learn to recognize common sources of bias.^33,40 While basic imaging informatics is already included in American Board of Radiology (ABR) core examinations as a non-interpretive skill, formal AI literacy and competency as an interpretive skill is a rising challenge. Trainee use of AI tools also needs to be carefully managed. There are benefits to learning how to use AI during residency but the timing of this exposure as well as the types of AI residents should use is open for debate. As residency training incorporates both interpretative skills (e.g., search patterns) and operational skills (e.g., communication with referrers), over-reliance on AI tools in either area could lead to deskilling, where radiologists lose essential skills due to limited clinical education, reduced case-based practice, and over-dependence on automated systems.

Radiology leaders should prepare their departmental infrastructure to be ready to meet the technical demands of emerging AI technology integration. Strategic investment decisions should prioritize flexible and scalable infrastructure, including hybrid- or cloud-based solutions that can efficiently support diverse AI tools with minimal downtime and security risk.⁴¹ As the market expands with more software offerings or vendors consolidate to reduce offerings, leaders have choices to make. Leaders should set up rigorous, transparent evaluation processes for systematically evaluating new AI software through pilot deployments, ongoing post-deployment performance monitoring, and clear governance structures to ensure accountability and alignment with patient care priorities and overall mission.⁴²

Pilot deployments should embed tools within real clinical workflows using clearly defined success and stopping criteria aligned with institutional priorities, and include participation by a representative group of users with accessible, transparent feedback mechanisms. Performance monitoring must extend beyond typical measures of diagnostic accuracy to relevant operational metrics like turnaround time, clinician satisfaction, and patient outcomes. This comprehensive approach is particularly important given that most studies of commercially available AI devices focus on technical performance, with limited existing evidence of real-world clinical and workflow impact.^43,44 Governance committees should include multi-disciplinary representation from across the enterprise including radiology, IT, legal, and other clinical departments depending on the application. These committees should establish institutional model registries tracking deployed AI tools and create formal escalation pathways for adverse events reporting. Recent institutional and multi-society endorsed frameworks provide practical guidance on AI deployment and regulatory implementation.^45,46

To enhance these evaluation and integration efforts, leaders should explicitly embrace principles of healthcare AI “delivery science,” a structured discipline focused on systematically designing, implementing, and evaluating healthcare innovations, as a means to practically translate AI technologies into sustained improvements in radiology care delivery.^47,48

Additionally, leaders should support structured collaboration and robust data-sharing frameworks, leveraging resources such as the American College of Radiology’s Assess-AI registry⁴⁹ with ACR Recognized Center for Healthcare-AI (ARCH-AI) institutional designation,⁵⁰ to contribute to national benchmarks and performance monitoring standards - similar to current practices in the ACR radiation dose registry⁵¹ and ACR Diagnostic Imaging Center of Excellence (DICOE) standards.⁵²

Ultimately, radiology as a specialty is facing pressure to adapt to exponentially increasing imaging volume which outstrips radiologist hiring, and necessity is the mother of invention. A possible innovative solution may come from advanced AI foundation models and autonomous agentic technologies.^53,54 These technologies, in parallel with emerging fields such as multi-omics and radiogenomics, may reposition radiologists as collaborative diagnostic specialists in multidisciplinary teams alongside pathologists, geneticists, and clinical data scientists. These require our specialty to address society-level fundamental sociopolitical, legal, and ethical concerns as health systems handle increasingly complex and sensitive multimodal data. Enterprises must familiarize themselves with evolving legal directives around these technologies.⁵⁵ On an individual level, radiologists must also stay aware of shifting perspectives regarding their own professional liability with increasing clinical AI integration.⁵⁶

Advanced Imaging Tools and Techniques

Overall Trends: Making Sense of the Hype Cycle

Advanced imaging tools and techniques encompass physical innovations in diagnostic acquisition as well as image-guided therapies (Figure 3, Table 2). Technologies that build upon familiar workflows, require minimal additional training, and have broader clinical applications have integrated rapidly, while those requiring greater operational lift with limited use cases are still undergoing deliberate evaluation.

Figure 3. The evolving role of AI in clinical workflows, demonstrated with an oncology CT staging example.

The left panel shows a continuum of increasing automation and decreasing human-in-the-loop involvement. This spectrum includes: (1) The physician working alone (Physician Primary); (2) The physician using narrow AI tools for discrete tasks while retaining all decision-making; (3) A foundation model executing multiple workflow components to support the physician; and (4) Single or multiple AI agents working autonomously with minimal human oversight (AI Model Primary). The right panel maps these AI paradigms to the clinical workflow of oncology staging, showing how narrow tools provide point solutions (e.g., nodule detection), a foundation model can handle a sequence of tasks (e.g., interpreting history and drafting a report), and a multi-agent system can orchestrate a complex, goal-oriented output by integrating various data sources and specialized AI capabilities.

Table 2.

Major applications in advanced imaging tools and techniques organized by the Gartner hype cycle. See footnotes for definitions.

Innovation Trigger	Peak of Inflated Expectations	Trough of Disillusionment	Slope of Enlightenment	Plateau of Productivity
Nanoparticle contrast agents1	PET-MRI3	Virtual reality7	HIFU (i.e. fibroid, bone)9	Dual energy CT12
Hyperpolarized 13C MRI2	Histotripsy4	Augmented reality8	Photon counting CT10	Theranostics13 - Iodine-131
	Ultra-high field MRI5		Theranostics11 - Lutathera	3D printing14
	Portable low field MRI6			Cinematic renderings15
				IR
				radioembolization16

¹Nanoparticle contrast agents: Tiny engineered particles used as contrast in imaging studies for enhanced disease detection and characterization.

²Hyperpolarized 13C MRI: Advanced MRI technique enhancing carbon-13 signals, significantly improving imaging of metabolic processes.

³PET-MRI: Combined imaging modality integrating positron emission tomography (PET) and magnetic resonance imaging (MRI) for simultaneous functional and anatomical insights.

⁴Histotripsy: Non-invasive ultrasound-based therapeutic technique using mechanical tissue fractionation for tumor ablation.

⁵Ultra-high field MRI: MRI scanners operating at very high magnetic field strengths (≥7 Tesla) providing enhanced spatial resolution and tissue contrast.

⁶Portable low-field MRI: Compact MRI scanners with lower magnetic field strength, enabling mobile, lower-cost imaging solutions.

⁷Virtual reality: Immersive, interactive 3D environments aiding in procedural planning, education, and patient engagement.

⁸Augmented reality: Overlaying digital imaging data directly onto real-world views to guide procedures or enhance image interpretation.

⁹High-Intensity Focused Ultrasound (HIFU): Targeted ultrasound treatment focusing high-energy beams for precise ablation of tissues, with the most common applications currently being fibroids and bone lesions.

¹⁰Photon-counting CT: Advanced CT technology employing detectors that count individual photons, improving resolution and reducing radiation exposure.

¹¹Theranostics – Lutathera: Radiopharmaceutical therapy combining imaging and targeted radiotherapy (Lutetium-177) primarily for neuroendocrine tumors.

¹²Dual-energy CT: CT imaging technique capturing images at two energy levels to better distinguish tissues and pathologies. Note there is variable traction in academic versus private practice settings.

¹³Theranostics – Iodine-131: Radioactive iodine used diagnostically and therapeutically, primarily for thyroid conditions.

¹⁴3D printing: Creation of physical models from imaging data, facilitating surgical planning, patient education, and medical training.

¹⁵Cinematic renderings: High-quality, realistic 3D visualizations generated from medical images to enhance anatomical comprehension.

¹⁶Interventional radiology (IR) radioembolization: Minimally invasive therapy delivering radiation-loaded microspheres directly into tumors through vascular catheters.

In imaging hardware, advanced imaging machines vary in their ability to progress beyond the academic setting to real-world practice. Adoption of these technologies has also been impacted by the steepness of the technical learning curve for radiologists and technologists. Dual-energy CT uses familiar clinical protocols while introducing new capabilities like spectral data and virtual images, and provides a familiar pathway for photon counting CT’s journey to adoption.^57,58 Portable low-field MRI potentially increases lower-cost access to MRI services for patients unable to travel to an MRI center, both locally and in low-resource countries.⁵⁹ However there is a tradeoff of lower image quality and clinical diagnostic value. Ultra-high-field MRI is typically used for neuroradiology applications at high-end academic centers but is limited by long acquisition times and limited additional clinical value.⁶⁰ These technologies are being adopted more slowly while their cost effectiveness and optimal applications are explored.^59–61 At the research end of the spectrum, highly specialized applications like hyperpolarized 13C MRI are limited to investigational settings.⁶² These newer innovations may also require radiologists to learn new interpretive skills, retrain technologists, and adjust workflows, which can further limit clinical translation. Each modality brings minimum competency requirements, ranging from spectral data interpretation for photon-counting CT to dosimetry calculations and radiopharmaceutical management for theranostics, that must be addressed through targeted training before widespread clinical adoption.

Advancements in diagnostic imaging devices have also led to parallel advancements in device-based therapeutic intervention capabilities. Visualization technologies such as 3D printing, physical models, and cinematic renderings provide intuitive 3D perspectives appreciated by referring clinicians.⁶³ Conversely, immersive technologies like augmented and virtual reality, despite potential to revolutionize image interpretation and procedural guidance, face adoption challenges due to necessary specialized equipment, training, and workflow modifications.⁶⁴ As these tools become more available and workflow-friendly, they could help position radiologists as key partners in procedural planning, transforming traditional static image interpretation into collaborative treatment designs. Maturation of diagnostic imaging tools has also helped shift a radiologist’s role from planning to treatment. For example, alongside progress in molecular imaging with new radiotracers and hybrid PET-MRI, there has been innovation in adapting use-cases for therapeutic applications with theranostics. Radiopharmaceuticals like radioactive iodine-131 or Lutathera (lutetium Lu 177 dotatate) are easily integrated into existing nuclear medicine workflows and benefit from extensive positive clinical trial data.⁶⁵ In comparison, device-based therapies like high-intensity focused ultrasound (HIFU) and histotripsy are still navigating safety and efficacy trials for different indications like liver tumors versus bone metastases.^66,67

Future-thinking: Strategies for Radiologists, Trainees, and Leaders

Given uneven adoption along the Gartner hype cycle of these devices and their respective techniques, it can be challenging to gain early experience, maintain expertise across more modalities, and commit resources to technologies with uncertain longevity.

Radiologists with a broadened skill set are needed as these advanced scanners and therapies settle into practice and demand increases, such as interpreting spectral CT maps, metabolic imaging, and ultra-field MRI scans. Depending on clinical referrer demand, there may be requirements to expand competency to radiopharmaceutical treatment and device-based ablation. The net effect may be a shift toward even more hyper-specialized radiologists; for example, neuro PET-MRI is already read at some centers by neuroradiologists cross-trained in PET, rather than split between nuclear medicine and neuroradiology.⁶⁸ Subspecialty-specific guidance is emerging to address unique imaging domains, such as recent multi-society recommendations for AI applications across the cardiac CT and MRI workflow.⁶⁹ Radiologists looking to expand their skill set can attend targeted webinars, seek hands-on demonstrations at conferences, learn from institutional or local experts, or enroll in dedicated training courses. Programs lacking certain advanced technologies should consider inviting guest speakers or visiting experts to share their experience and conduct workshops if feasible.

Trainees have recognizably unequal access to advanced imaging techniques, including both local opportunities restricted by training level, or not having availability locally at all. A complete lack of access embodies institutional privilege and creates inequalities in training that can then perpetuate throughout a career, ultimately affecting patient access to advanced imaging services. When techniques are restricted to attendings or fellows, residents should proactively request dedicated electives or explore research opportunities. The dual certification in nuclear medicine pathway may allow residents to get experience with theranostics that would be otherwise unavailable.⁷⁰ Fellowship selection should specifically target institutions with robust advanced imaging programs that address identified gaps in residency training, positioning trainees for the increasingly specialized roles that these technologies demand.

Given the substantial financial and training investments required for many of these advanced imaging tools, radiology leaders must ensure adoption decisions align with institutional strengths and clinical demand from referring specialties. For example, establishing a histotripsy program makes much more sense for a specialized cancer center than a community hospital. They should proactively invest in infrastructure upgrades, including data storage expansion for multimodal PET-MRI imaging and specialized facilities like hot labs for theranostics programs. AI could itself potentially help by anticipating these infrastructure needs through the use of predictive models and optimization algorithms that inform facility design, room layout, and equipment placement. Leaders should also establish formal steering committees with representatives from different sections to evaluate emerging modalities and coordinate with clinical departments to identify institutional priorities.

Overall, uneven adoption patterns of advanced imaging technologies are worsening inequalities in access to healthcare training and expertise. Institutions with greater resources can offer cutting-edge tools while others lag behind, creating disparities in diagnostic and therapeutic options based on geography and thereby harming patient care. These disparities manifest across multiple dimensions, such as academic medical centers versus community hospitals, or resource-rich versus resource-poor regions globally. Vendors seeking broader adoption should recognize that gatekeeping advanced technologies slows dissemination and limits the pipeline of trained radiologists, and ideally should collaborate with radiology leaders to make technologies available to more trainees or implement subsidized cloud-based deployment models that reduce financial requirements for resource-limited settings, thus expanding patient access as well. Radiology as a field must discuss the ethical implications of committing substantial resources to technologies with uncertain longevity while basic imaging needs remain unmet in underserved areas. Adoption decisions should balance innovation with equitable access to ensure imaging advances strengthen rather than fragment patient care.

Clinical Practice Paradigms

Overall Trends: Making Sense of the Hype Cycle

Clinical practice paradigms affect how radiology practices function operationally as integrated systems, influencing workflows at all levels from radiology technologists to administrators (Figure 4, Table 3). These paradigms have arisen in response to increasing workforce demands and background sociotechnical developments, with innovations spanning AI-guided ultrasound scanning to mature teleradiology practices.⁷¹

Figure 4.

The growth of advanced imaging tools and techniques can be categorized into four key areas: Emerging Imaging Modalities, Visualization and Workflow Tools, Theranostics and Molecular Imaging, and Image-Guided Interventions.

Table 3.

Major clinical practice paradigms organized by the Gartner hype cycle. See footnotes for definitions.

Innovation Trigger	Peak of Inflated Expectations	Trough of Disillusionment	Slope of Enlightenment	Plateau of Productivity
Robotics1	Edge computing4	AI-guided workflows6 - Scheduling & No show prediction	Multi-dose contrast injectors9	Cloud-based PACS13
AI-guided workflows2 - Exam protocoling	AI-based quality assurance (QA)5	Patient portals for imaging results7	Workflow orchestrator10	Teleradiology14
AI-guided technologist workflow3		Radiology assistants8	Enterprise imaging platforms11	Predictive scanner maintenance15
Ultrasound scanning
			Remote scanning12	Independent office-based labs for IR16
				Radiation dose management software17

¹Robotics: Very conceptual and early-stage; potential roles include remote procedural assistance, robotic-assisted interventions, and possibly robotics for scanner positioning or patient care.

²AI-guided exam protocoling: Use of AI to automate the customization of imaging protocols based on clinical indication and patient data.

³AI-guided technologist workflow/Ultrasound scanning: AI assistance to technologists for optimized scanning protocols and real-time guidance during ultrasound examinations.

⁴Edge computing: Data processing performed directly at the point-of-imaging (i.e. scanner level) to improve speed and reduce network load.

⁵AI-based quality assurance (QA): Tools automatically assessing imaging quality in real-time or retrospectively, reducing repeat scans and improving diagnostic consistency.

⁶Patient portals for imaging results: Secure online platforms allowing patients direct, easy access to their imaging studies and results.

⁷Radiology assistants: Advanced-practice radiology technologists who support radiologists by performing pre-determined clinical tasks, patient interactions, and some procedural activities.

⁸AI-guided workflows – Scheduling & no-show prediction: Algorithms optimizing scheduling processes and predicting appointment attendance to improve resource utilization.

⁹Multi-dose contrast injectors: Systems enabling precise, automated delivery of contrast media for multiple patients, optimizing workflow efficiency, reducing waste, and standardizing dosing.

¹⁰Workflow orchestrator: Integrated software coordinating and optimizing radiology department workflows to enhance efficiency and reduce operational bottlenecks.

¹¹Enterprise imaging platforms: Centralized systems consolidating imaging data across departments or institutions, facilitating better management, access, and integration.

¹²Remote scanning: Technologies enabling radiology technologists or radiologists to operate imaging equipment from remote locations to increase accessibility.

¹³Cloud-based PACS (Picture Archiving and Communication System): Cloud-based infrastructure for storing, accessing, and managing imaging data, facilitating scalable and remote access.

¹⁴Teleradiology: A clinical practice model enabling distributed radiology workflows, allowing radiologists to remotely interpret images and deliver timely diagnostic services irrespective of geographic or temporal constraints.

¹⁵Predictive scanner maintenance: AI-driven systems proactively monitoring equipment performance, predicting potential failures, and scheduling preventive maintenance.

¹⁶Independent office-based labs for interventional radiology: Freestanding clinical facilities where interventional radiology (IR) procedures are performed outside traditional hospital settings.

¹⁷Radiation dose management software: Systems tracking and optimizing radiation doses, enhancing patient safety and regulatory compliance.

An initial shift has already transformed radiology practice by focusing on department-wide connectivity and data management. On the horizon is edge computing, an approach to move data processing to gateways within the imaging suite itself rather than a centralized server, thereby minimizing network load and bandwidth demands.⁷² By processing imaging data at the scanner level, edge computing enables real-time AI applications such as automated patient positioning, protocol adjustments, and immediate quality verification.⁷³ Cloud-based PACS offering storage and processing scalability with remote access, and enterprise imaging platforms that consolidate imaging data have fundamentally reshaped how radiology departments operate, allowing 24/7 teleradiology coverage.^74,75 Although teleradiology jobs have existed for many years, adoption of remote work spurred during the COVID-19 pandemic has continued its upward trend even in academic environments.^76,77 Early evidence suggests teleradiology can maintain or improve operational metrics, with 119 of 124 radiologists in one study reporting similar or improved report turnaround times when working remotely, though comprehensive data on long-term workforce satisfaction and quality outcomes remain limited.⁷⁸

Operationally, this has also affected radiology technologists and the overall workflow of radiology practices. Remote scanning technologies are already allowing technologists to operate imaging equipment from distant locations. Early implementations have demonstrated potential for addressing critical technologist shortages, particularly in rural and underserved areas. However, successful deployment requires robust telecommunications infrastructure, standardized protocols for patient safety and positioning verification, and clear delineation of on-site versus remote responsibilities.^79,80 On the other hand, more nascent technologies such as AI guidance for optimizing ultrasound scanning or automating exam protocoling are just being developed with many remaining technical hurdles before widespread adoption can be considered.^81–83 Robotics, similarly at an early stage, is being explored for automated patient positioning, precision-guided interventional procedures, and enhancing remote operational functionalities (i.e. robot-performed ultrasound), although its true practical utility is too early to predict.^84–86 These combined advances in remote capabilities and task-specific automation will help address critical technologist staffing shortages and improve imaging access in rural and underserved communities.

Another advancement in radiology clinical practice targets specific high-value tasks within established workflows, such as efforts to optimize scanner and network use through predictive analytics and infrastructure improvements. Scanner manufacturers now offer proprietary systems that continuously monitor scanner health and flag component issues early, minimizing inadvertent downtime.^87,88 AI has been explored to better schedule patients to optimize scanner utilization; however, as a solution its effectiveness may be more limited by physical room constraints or transport workflow bottlenecks specific to each setting rather than no-show prediction, and radiologists or practices that do not own equipment may find limited utility.⁸⁹ With supply chain disruptions and concerns about waste, adaptations such as multi-dose injectors and low-dose imaging techniques are becoming a growing area of innovation.^90,91 As these non-interpretive patient care delivery improvement tools mature, they may help radiology departments meet the surging demand for imaging.

Future-thinking: Strategies for Radiologists, Trainees, and Leaders

Innovations in radiology practice paradigms can be more nebulous to define and recognize, however these evolutions directly set trajectories for quality, education, and operational standards and affect patient care.

Practicing radiologists should thoughtfully consider how evolving paradigms could impact their daily clinical roles, academic involvement, and team interactions.⁹² Overall, teleradiology has changed the modes of potential communication but has not reduced its necessity or importance. Teleradiology’s advantages (e.g., saved commuting time, flexible location and work hours, enhanced productivity) also share risks of diminished organizational engagement, collaboration, and professional visibility. These challenges can be partially addressed through intentional professional communication practices: virtual multidisciplinary conferences, formal mentorships, and timely feedback to technologists and trainees alike.⁷⁷ With the 21st Century Cures Act raising patient expectations for rapid transparent access to imaging results, there has been increasing adoption of patient portals, putting pressure on radiologists to also consider innovations in patient-centered communication.^93,94 Within this context of increasingly remotely-practicing radiologists, another growing trend to watch is the broader integration of advanced practice roles, including nurse practitioners (NPs), physician assistants (PAs), and registered radiologist assistants (RRAs), who can enhance patient care by performing selected procedures and providing preliminary assessments under radiologist supervision, however legitimate concerns exist regarding potential scope creep over time and its implications for quality of care and outcomes.⁹⁵.⁹⁶

Trainees must actively consider how emerging clinical practice paradigms will shape their training experiences and future professional identities. While virtual and hybrid reading setups offer greater exposure to multiple subspecialty attendings, flexible participation in teaching sessions, and reduced travel between sites, they also create challenges including reduced hands-on teaching, isolation, less spontaneous mentorship, and lack of timely feedback.^{97 77} For these reasons, trainees often support a combination of virtual and in-person readouts.⁹⁸ Training programs need to implement intentional strategies including virtual case reviews, scheduled one-on-one membership sessions, and clear expectations for structured feedback - to preserve effective learning experiences.⁹⁹ As trainees transition to independence and select their first jobs, they will need to carefully evaluate how “future-ready” a prospective practice may be, including the practice’s responsiveness to flexible work, readiness for technological innovations, and support for professional development for maintaining skills or up-skilling. Importantly, trainees are not just passive participants in practice culture but will be instrumental in shaping it. Given trainees’ anticipated longevity in the profession, with their greater expected cumulative exposure to technological innovation, and on-the-ground cross-subspecialty clinical experiences, they are uniquely positioned to bridge generational gaps in understanding and integrating technologies like AI within broader practice contexts. It is important to recognize that trainees, faculty, mentors, and peers across different generations (i.e. Gen Z and Alpha) will likely experience and navigate AI and innovation differently, so trainees should intentionally cultivate cross-generational communication skills.¹⁰⁰ As both learners and educators, trainees can facilitate multi-directional knowledge sharing to drive innovation that remains relevant across all levels of our profession.

Radiology leaders face new radiology workforce dynamics and challenges: flexible radiologist staffing models and distributed work arrangements, sheer breadth of technology innovations, and preservation of broader practice culture (or academic mission) with increased clinical demands. To recruit and retain radiologists in diverse practice settings amidst workforce shortages, leaders should reevaluate recruitment approaches, faculty development programs, and career incentives, and also potentially advocate for policies for International Medical Graduates (IMGs) inclusion.^101–103 Also, while employing more RRAs to assist with overloaded clinical tasks is under consideration, doing so could exacerbate radiology technologist shortages as being a technologist is a prerequisite for becoming an RRA.¹⁰⁴ Additionally, new challenges arise when integrating tools such as predictive analytics for equipment maintenance and scheduling, with decisions needing to be made on how autonomous these systems should be - to minimize operational disruptions and preserve patient safety.^105,106 Radiology leaders now must also either seek experts in or become personally fluent at the trifecta to make good decisions: clinical operations, academic and scientific mission, and radiology informatics, as there are increasing demands on time with influxes of new technologies (“shiny object syndrome”).

We must thoughtfully navigate the innovation hype cycle toward a broader, multidimensional view of sustainable radiology practice paradigms. As we look towards the future, cross-dimensional challenges - such as provider burnout, cybersecurity threats, or the growing importance of environmental sustainability - will increasingly shape innovations in clinical practice paradigms. Burnout poses substantial risks to workforce resilience, potentially undermining care quality and long-term individual and overall practice sustainability - and artificial intelligence may worsen this phenomenon.^107–109 Addressing this burnout requires holistic, systemic solutions rather than isolated interventions - and also careful consideration of how new innovations impact people and organizations. Similarly, cybersecurity should evolve from focusing only on discrete data repositories and networks, to more holistic strategies as the integration of complex AI applications throughout the broader patient lifecycle has increased vulnerabilities across the entire digital healthcare ecosystem.^110,111 Security should also consider alignment. This emphasis on systemic, interconnected thinking is equally critical for addressing global environmental sustainability challenges and planetary health, which require deeper thinking around the resource implications of our increasingly technology-driven workflows, from imaging equipment choices to energy-efficient computational tools and facilities. The concept of sustainability also inherently extends further, and encompasses social equity, workforce resilience, economic viability, and adaptability to unforeseen global disruptions (such as pandemics, supply chain interruptions, tariffs, workforce shortages, etc.)^41,112,113 As the world evolves, paradigms of radiology practice evolve with it, and proactively embracing this comprehensive multidimensional outlook will better position radiology practices to deliver consistent, high-quality care amidst ongoing technological, environmental, and societal shifts.

Conclusion

The Gartner hype cycle and its parallel technology adoption lifecycle helps reveal patterns among technologies to identify potential trends and trajectories for future-thinking. When radiology technologies are more mature (software, hardware, or larger practice paradigms), they more readily demonstrate clinical and operational value, compared to technologies at earlier stages which often have more limited clinical scope, need more robust evidence, or face major implementation barriers. Nevertheless, the hype cycle is not an end-all, be-all tool. It tries to capture technology maturity, but has gaps in describing true market and clinical uptake for a given setting; for example dual-energy CT in academic versus private practice settings, on account of factors like varied institutional priorities or workflow setups.

With the rapid proliferation of radiology innovations, departments must establish clear processes for deciding what technologies merit investment - a “go or no-go” framework - to systematically evaluate clinical utility, operational feasibility, scalability, and return on investment. Practices must look at their institutional strengths and the needs of their referral base to decide what tools will benefit them the most. These governing frameworks should explicitly consider the limitations of tools like the Gartner hype cycle, recognizing that technology innovation maturity does not inherently guarantee local appropriateness or feasibility. Leaders should be able to determine precisely when to implement, pause, or forego innovations based on institutional priorities, workflow compatibility, and clear expected value.

We propose a systematic approach to technology adoption that maps innovations to their maturity levels, evaluates them through a quantitative five-domain decision framework (Figure 6 ), and implements them using delivery science principles to maximize clinical value.⁴²

Figure 6. “Go- or No-go” Conceptual Framework for Radiology Technology Adoption.

A systemic framework for radiology technology adoption that evaluates innovations across five domains (clinical impact, operational feasibility, financial viability, institutional alignment, ethical and equitable considerations) to guide decision making with three distinct options: go, pause, no-go.

Radiology in 2035 and beyond will look much different than it does today (Table 4), as AI becomes embedded throughout the radiology cycle, integrated multi-omics drives a shift to multidisciplinary radiology, radiologists adopt advanced diagnostic and therapeutic skill sets, and practice environments continue to evolve. We must begin preparing for this transformation now by upgrading infrastructure to meet rising technical requirements, adapting training to incorporate new technologies, and engaging in societal-level legal and ethical discussions revolving around evolutions in patient care delivery. Additionally, radiology as a field is facing the opportunity to proactively shape the creation and adoption of emerging technologies, ensuring that innovations align closely with clinical needs and values from the outset.¹¹⁴ Long-term success will hinge on meeting the moment, facing cross-cutting challenges (such as cybersecurity, sustainability discussed before), and adapting to a rapidly changing global landscape.

Table 4.

10 predictions for radiology’s future (2025–2035). This 10-prediction summary was inspired by Dr. Curtis Langlotz’s predictions for the future of AI and informatics from 2023,³¹ with additional influence by our above analysis of the Gartner Hype Cycle framework.

Category	Prediction
Software and Algorithms	Foundation models will be routinely used across radiology workflows
	Autonomous AI agents will be deployed for complex or collaborative tasks like tumor boards or patient communication
	Natural language will become a more universal imaging & reporting interface
	Quantum computing will grow as a promising research area
	Clinical AI regulatory frameworks will evolve from static approval to continuous learning systems
Advanced Imaging Tools and Techniques	AI will transform routine imaging into comprehensive opportunistic screening platforms
	Theranostics will become a primary treatment pathway
Clinical Practice Paradigms	Radiology will evolve as a quantitative diagnostic specialty, central to multi-omics teams for precision medicine
	Remote technologist networks will solve critical staffing shortages through AI-guided scanning
	Residency training and board certification will fundamentally incorporate AI competency

At the same time, as these technologies mature, there is a risk of unequal adoption resulting in systemic inequality where innovations are concentrated in select, resource-rich institutions or practices. This future possibility threatens to create a tiered-healthcare system where patient access to advanced technology is increasingly limited by geographic and socioeconomic factors. Innovation rollout ideally should be done thoughtfully and equitably, however the realities of translational deployment suggests that policymakers, researchers, and health advocates should evaluate best practices for health equity. Addressing these disparities requires collaborative infrastructure development including shared registries for performance monitoring (such as ACR’s Assess-AI and ARCH-AI programs), multi-site validation to establish real-world benchmarks across diverse practice settings, and equitable access initiatives that ensure technical innovations benefit the entire radiology ecosystem rather than just high-resource institutions. Also, a potential underexplored avenue is that vendors can help by partnering with academic programs to increase innovative technology availability to trainees, which in turn can increase patient access while also accelerating technological adoption.

Time will test radiology’s ability to adopt and implement these transformative technologies systematically. How well the specialty navigates this transition will determine whether emerging innovations fulfill their promise of improved patient care or fade into obscurity - as new examples of unfulfilled hype.

Figure 5. Clinical practice paradigms: workflow and operational innovations.

This diagram illustrates innovations in clinical practice paradigms in radiology, organized into a three-stage linear workflow: (1) pre-imaging and scheduling, (2) imaging acquisition and support, and (3) post-imaging and communication (blue). These operational workflows are underpinned by (4) integrated operational management and practice models, and (5) and robust infrastructure components (green). Surrounding these paradigms are overarching exemplar considerations of burnout, cybersecurity, and sustainability (orange), highlighting their cross-cutting impact across potential developments in radiology practice paradigms.