AI in Radiology: Navigating the New Regulatory Landscape
"How the FDA's innovative approach to AI regulation is reshaping the future of medical imaging and diagnostics."
Artificial intelligence (AI) is rapidly transforming healthcare, and radiology is at the forefront of this revolution. As AI-powered tools become increasingly prevalent in medical imaging, radiologists are keen on understanding the implications of these advancements. The AI health market is experiencing explosive growth, projected to surge from $600 million in 2014 to $6.6 billion by 2021. This growth underscores the urgent need for clear regulatory guidelines to ensure the safe and effective integration of AI into clinical practice.
Radiologists are particularly invested in AI's development because many significant breakthroughs have occurred in image-based technologies. However, the regulatory landscape is still evolving, and the FDA's approach to overseeing AI development is taking shape. Balancing innovation with patient safety remains a key challenge.
While the FDA has decades of experience regulating drug development, evaluating AI technologies requires a new framework. Recently, the FDA has started focusing on this challenge and is exploring innovative programs to handle AI's unique aspects. This article will delve into the FDA's current strategies for regulating AI in radiology, including the Digital Health Software Precertification Program (Pre-Cert), and discuss the potential impact on radiologists and patient care.
Radiology’s Share of FDA-Cleared AI
A 2026 roundup reports that the AI medical-imaging market is valued at $3.85 billion and that more than 950 AI devices have FDA clearance, with 76.8% in radiology; it also cites 20% higher cancer detection, 38 minutes saved in stroke triage, and AI adoption at 48% of U.S. hospitals. A separate analysis characterized 1,094 radiology AI devices authorized through December 2025 and found annual authorizations rose 42% over the period studied. Another report gives radiology’s share as 76% of 1,430 FDA-authorized AI/ML devices through December 2025. The reported radiology shares are similar but not identical, and the sources describe different device counts and analyses.
FDA Review and Implementation Challenges
FDA review of radiologic AI includes the agency’s processes for software as a medical device and regulatory designs specific to radiology machine-learning algorithms, with the aim of supporting safety in clinical practice. A 2024 review also discusses challenges that arise when implementing these tools and maximizing their efficacy. In June 2026, the FDA adopted a classification and predicate for radiological machine-learning-based quantitative imaging software with a predetermined change control plan (PCCP); the source describes it as the first device classification incorporating PCCPs into special controls. A 2025 systematic review examines FDA premarket clearances for radiology AI/ML devices, with emphasis on testing.
From Pre-Cert to a Broad Device Cohort
The FDA’s Digital Health Software Precertification Pilot Program explored innovative approaches to regulatory oversight of medical-device software. Later work describes a field with more than one thousand FDA-authorized radiology AI devices and characterizes the devices by function, clinical application, manufacturer segment, and regulatory features. Together, these sources mark a shift from exploring oversight approaches for software to examining the characteristics of a large authorized-device cohort. The supplied material does not provide further details about the pilot’s outcomes or the cohort’s specific findings.
The FDA's Evolving Regulatory Approach
Traditionally, the FDA approves medical devices, including AI-based ones, through pathways like the de novo premarket review. This pathway is designed for new devices that pose low to moderate risk and lack a legally marketed predicate device. Between February and May 2018, the FDA approved three AI-based devices for stroke, diabetic retinopathy, and wrist fracture using this route. This demonstrates the FDA's initial steps in addressing AI in healthcare.
- Faster Approval: Unlike traditional processes that focus on specific products, Pre-Cert provides a pathway for approving entire organizations.
- Organizational Focus: Once an organization is Pre-Cert approved, it can bring lower-risk products to market without premarket review.
- Flexibility: The FDA suggests that Pre-Cert may allow firms to make minor device changes without requiring new submissions each time.
Regulation, Acquisition, and Workflow
A mini-review on AI in anatomic pathology says commercial AI products must obtain marketing approval, such as Conformité Européenne approval or FDA clearance, before they can be legally marketed for clinical use. It identifies successful integration into pathologists’ workflow as the ultimate test for a product. A radiology-focused article notes that addressing regulation, security, and privacy when acquiring AI solutions is complex. These sources highlight that authorization and practical workflow fit are distinct considerations.
Oversight and Liability in AI Workflows
A 2026 article examines how FDA-cleared AI-assisted radiology workflows affect liability, drift monitoring, and radiologist oversight in hospitals. Its supplied summary identifies these as connected issues in the use of cleared tools. The source does not provide comparative findings or specific liability rules, so no conclusions about how liability is allocated can be drawn from the available material.
Challenges and the Path Forward
While the Pre-Cert program offers a promising approach, it also raises important questions. Vetting organizations instead of individual products is unprecedented, creating concerns about monitoring safety and efficacy. The FDA is actively working on defining different levels of certification and ensuring that organizations maintain their Pre-Cert status over time.
Pre-Cert and Software Market Entry
In April 2018, FDA commissioner Dr. Scott Gottlieb announced development of the Digital Health Software Precertification Program, also known as Pre-Cert. The program sought to streamline regulation of software technology, including AI products. The source presents Pre-Cert as a new program intended to change how software products come to market.
Uneven Adoption and Expanding Technologies
A 2025 overview reports that hundreds of AI-enabled tools had received regulatory clearance for medical-imaging tasks by late 2025. It describes clinician adoption as growing but uneven around the world. A second 2025 overview covers FDA authorizations, clinical adoption rates, and technologies ranging from convolutional neural networks to foundation models for medical imaging. The supplied summaries do not give specific adoption rates or forecasts.
Planned Changes Through PCCPs
A predetermined change control plan (PCCP) describes anticipated device modifications, the methodology for developing, validating, and implementing them, and an assessment of their impact. PCCPs can apply to radiology-specific devices. The source says they are frequently used to retrain or optimize AI algorithms with additional datasets. This offers a defined way to describe planned changes to an AI device.
Radiologists’ Knowledge and Attitudes
A cross-sectional study examines Norwegian radiologists’ knowledge and attitudes toward artificial intelligence. Its supplied summary states that AI solutions have been approved and adopted in radiology. The summary provides no results about the radiologists’ specific knowledge, attitudes, or experiences, so those findings cannot be characterized from the available material.
Establishing a robust framework for risk stratification is critical. The FDA plans to use the International Medical Device Regulators Forum framework, which categorizes risk based on the patient's situation (critical, serious, or nonserious) and how the device's information will be used (inform clinical management, drive clinical management, or treat and diagnose). However, real-world implementation may prove challenging due to the lack of established precedents.
Some healthcare experts worry that the Pre-Cert approach could give software an implicit stamp of approval without sufficient clinical validation. As the FDA continues to refine its regulatory strategies, radiologists must stay informed and actively participate in shaping these changes. The first official version of the Pre-Cert program (Pre-Cert 1.0) is expected in December 2018, necessitating ongoing collaboration with the FDA to ensure a safe, effective, and patient-centered approach to AI in radiology.