America’s Largest Hospital System Ready to Start Replacing Radiologists With AI
A debate over the role of artificial intelligence in healthcare has intensified after Mitchell Katz suggested that AI could replace a significant portion of radiologists if regulatory barriers are removed. He argued that AI systems could handle initial tasks like reading mammograms and X-rays, potentially reducing costs and improving efficiency in hospitals. However, the remarks triggered strong backlash from medical professionals, who warned that relying solely on AI could risk patient safety and lead to misdiagnosis. Experts emphasize that while AI can assist in diagnostics, human oversight remains essential, highlighting the ongoing tension between innovation and safety in the evolving healthcare landscape.
A fresh debate over the role of artificial intelligence in healthcare has emerged after the head of one of the United States’ largest public hospital systems suggested that AI could replace a significant portion of radiologists.
NYC Health + Hospitals CEO Mitchell Katz said during a panel discussion that hospitals could already substitute many radiology tasks with AI systems if regulatory hurdles were addressed.
AI as a Cost-Saving Tool
Speaking at an event hosted by Crain’s New York Business, Katz indicated that AI-powered tools could streamline diagnostic processes and reduce operational costs.
“We could replace a great deal of radiologists with AI at this moment, if we are ready to do the regulatory challenge,” he said.
He pointed to areas such as breast cancer screening, where AI systems could initially analyze scans and flag abnormalities, potentially reducing the need for human intervention in routine cases. According to Katz, such an approach could lead to “major savings” for hospital systems.
Strong Pushback From Medical Community
The remarks quickly drew criticism from medical professionals, particularly radiologists who questioned both the safety and feasibility of such a transition.
Mohammed Suhail, a radiologist based in San Diego, warned that relying solely on AI for medical imaging could put patients at serious risk.
He described the idea as dangerous, arguing that current AI systems are not capable of replacing expert human judgment in complex diagnostic scenarios. According to Suhail, any attempt to fully automate radiology readings could lead to misdiagnosis and potentially severe consequences for patients.
Concerns Over AI Reliability in Medical Imaging
The controversy is further intensified by emerging research highlighting limitations in AI diagnostic tools.
A recent study by researchers at Stanford University found that some AI models used for analyzing chest X-rays could perform well on benchmark tests even without access to actual medical images. Instead, the systems generated convincing but unfounded explanations, a phenomenon researchers described as an “AI mirage.”
Unlike typical AI errors or “hallucinations,” these outputs appear coherent and medically plausible, making them harder to detect and potentially more dangerous in clinical settings.
Balancing Innovation and Patient Safety
The discussion underscores a growing tension in healthcare: balancing the efficiency and cost benefits of AI with the need for accuracy and patient safety.
While AI tools are increasingly being used to assist radiologists such as prioritizing urgent cases or highlighting anomalies experts caution against removing human oversight entirely.
For now, regulatory approval, clinical validation and ethical considerations remain significant barriers to widespread AI replacement of medical specialists.
A Broader Industry Shift
Katz’s comments reflect a wider trend across industries where organizations are exploring AI to improve efficiency and reduce costs. However, in high-stakes fields like healthcare, the margin for error remains extremely small.
As hospitals and policymakers evaluate the future role of AI in diagnostics, the debate is likely to intensify particularly as new research continues to reveal both the promise and the risks of advanced AI systems.
