AI Algorithms in Radiology Face Skepticism Despite FDA Approvals

The article discusses the slow adoption of AI-powered algorithms in radiology, despite FDA approval of over 700 algorithms, due to concerns over reliability, transparency, and demographic representation. Radiologists are hesitant to implement AI technology, citing limited testing and uncertainty about its effectiveness in real-world settings, which could hinder progress in disease detection and patient outcomes." This description focuses on the primary topic of AI adoption in radiology, the main entities involved (radiologists, FDA, and AI algorithms), and the context of disease detection and patient outcomes. It also highlights the significant actions and consequences, such as the slow adoption and potential hindrance to progress, and provides objective and relevant details that will guide the AI in creating an accurate visual representation of the article's content.

Bijay Laxmi
New Update
AI Algorithms in Radiology Face Skepticism Despite FDA Approvals

AI Algorithms in Radiology Face Skepticism Despite FDA Approvals

Over 700 AI-powered algorithms have been approved by the FDA to aid radiologists in detecting diseases, with more than 75 of them specifically designed for radiology. However, only 2% of radiology practices are currently using AI technology, citing limited testing, lack of transparency, and demographic concerns.

Why this matters: The adoption of AI technology in radiology has the potential to significantly improve disease detection and patient outcomes, but the slow uptake by radiologists could hinder progress in healthcare. Moreover, the effectiveness of AI algorithms in real-world settings has broader implications for the development and regulation of AI in medicine.

Radiologists remain hesitant to adopt AI technology due to questions about the reliability of the algorithms in real-world settings. Dr. Curtis Langlotz, a radiologist at Stanford University, notes, "If we don't know on what cases the AI was tested or whether those cases are similar to the kinds of patients we see in our practice, there's just a question in everyone's mind as to whether these are going to work for us."

Despite the potential benefits of AI in improving accuracy and efficiency, the culture of medicine has been slow to embrace the technology. Dr. Ronald Summers, a radiologist and AI researcher at the National Institutes of Health, believes that some AI techniques are so good that they should be implemented now, questioning why valuable information is being left unused.

In 2022, European regulators approved the first fully automatic software that reviews and writes reports for chest X-rays that appear healthy and normal. The company behind the app, Oxipit, is currently submitting its application to the FDA for approval in the United States.

Research has shown that radiologists tend to overestimate their own accuracy. A study by Koios Medical found that physicians viewing the same breast scans disagreed with each other more than 30% of the time on whether to perform a biopsy. The same radiologists even disagreed with their own initial assessments 20% of the time when viewing the same images a month later. According to the National Cancer Institute, about 20% of breast cancers are missed during routine mammograms.

AI-powered algorithms are currently being used to assist radiologists rather than replace them. Dr. Laurie Margolies of Mount Sinai hospital system in New York tells patients, "I looked at it and the computer looked at it, and we both agree." She believes that this dual agreement between the radiologist and AI gives patients an even greater level of confidence in the results.

Experts predict that AI will continue to work alongside radiologists, functioning like autopilot systems on planes by performing important navigation tasks while always remaining under the supervision of a human pilot. Initial results from a Swedish study of 80,000 women showed that a single radiologist working with AI detected 20 more cancers among mammograms than two radiologists working without the technology.

As AI algorithms continue to advance, the adaptation of the health technology assessment (HTA) process through a methodological framework for AI-based medical devices can provide valuable insights into their effectiveness, cost-effectiveness, and societal impact. This enhanced evaluation process aims to guide the responsible implementation of AI in radiology and maximize its benefits for patients and healthcare systems. However, much uncertainty remains regarding the reliability of AI algorithms, data quality, and regulatory processes, presenting multiple challenges for HTA agencies in assessing and approving these new technologies.

Key Takeaways

  • Over 700 AI-powered algorithms approved by FDA for radiology, but only 2% of practices use them.
  • Radiologists hesitant due to limited testing, lack of transparency, and demographic concerns.
  • AI can improve disease detection and patient outcomes, but slow adoption hinders progress.
  • AI-powered algorithms currently assist radiologists, rather than replace them, to improve accuracy.
  • Enhanced evaluation process needed to guide responsible implementation of AI in radiology.