The Dominance of AI in Radiology: Insights from a New Study by GigHz

In a groundbreaking study led by Dr. Pouyan Golshani, founder of GigHz, findings reveal a significant trend in the authorization of artificial intelligence (AI) medical devices. The research, published in the journal Cureus, indicates that a remarkable 76.5% of the 1,430 FDA-reviewed AI medical devices have been authorized specifically in the field of radiology. This comprehensive analysis raises an essential question for healthcare systems investing in these emerging technologies—what truly constitutes a clinical workflow that is prepared to integrate AI?

Overview of the Study

The peer-reviewed study evaluates authorization records dating from September 1995 to December 2025. It highlights that the radiology panel alone accounted for 1,094 of the recorded devices, with cardiology and neurology panels together making up an additional 14.1%. Together, these three panels amounted to an impressive 90.6% of all authorizations examined. These statistics indicate not just an interest in AI within the healthcare field, but also a concentrated effort and existing infrastructure that already supports digital medical imaging in radiology.

Dr. Golshani attributes this concentration to radiology's established digital systems, which facilitate the processing and analysis of medical images. He states, “Radiology already had digital images, common file standards, and systems that effectively move scans to the person interpreting them. This creates a conducive environment for the development of AI technologies.” However, he cautions against the misconception that the integration of AI means an impending automation of a radiologist’s role. The complexity and nuance involved in radiology practice remain substantial.

A Surge in Authorizations

Examining the growth over time, the study shows that annual authorizations averaged only 1.8 devices per year from 1995 through 2014. A dramatic increase is evident, with an average of 264 authorizations per year from 2023 through 2025. Despite the overall growth, the study also highlights a trend where many companies remain specialized, with 67.8% of developers having only a single authorized AI device. Conversely, a small group of developed firms account for the preponderance of devices.

Limitation of Specialties

Interestingly, the gap in AI device authorizations across specialties is noteworthy. For instance, very few authorizations were documented under categories like pathology, microbiology, and obstetrics/gynecology—highlighting a potential disparity in the application and acceptance of AI technologies in various fields.
While various specialties could benefit immensely from AI-driven research and implementations, the findings point to an existing ecosystem that supports radiology more robustly than others, demonstrating the need for focused development in areas that lack similar technological infrastructure.

The Path Forward

Dr. Golshani emphasizes the need for a balanced approach amid the rapid growth of AI technologies in healthcare, suggesting that developers and health systems should establish clear clinical tasks paired with necessary data accessibility. He argues that it's important to assess tools based on how they enhance clinical decision-making, not merely upon their existence. “The challenge is ensuring that the crucial information is gathered and structured appropriately, so a physician can access it in a timely manner during consultations,” he adds.

In addition, assessing the consequences of poorly thought-out AI integration—fear of both obsolescence and missing out—could lead healthcare entities to make hasty decisions. Golshani challenges us all to be mindful of the tool’s effectiveness, identifying areas where it may fall short, and defining responsible accountability for any failures.

Understanding the Study's Scope

It is essential to note that this analysis is focused on FDA authorization records, which does not equate to clinical adoption or patient benefits. The FDA states that its AI-enabled device list isn't exhaustive of all healthcare AI applications, reflecting the nascent stage of technological development in specialties beyond radiology. Dr. Golshani has also authored a policy brief addressing how state oversight in relation to clinical decision-support software connects with federal regulatory practices.

In summary, as AI technologies continue to evolve and permeate the healthcare landscape, understanding the underlying factors influencing their integration into clinical workflows will be paramount for effective patient-centered care. It is evident that as the field progresses, there remains much to explore and redefine within the realm of medical technology.