FDA Keeps Radiology AI Revenue Tied to Premarket Clearance

The Food and Drug Administration (FDA) has extended its strict regulatory oversight over artificial intelligence-driven radiology software, preventing new AI algorithms from directly generating revenue in the United States. This decision, effective September 17, reinstates the agency's April 1 denial of a petition from Australian healthcare AI company Harrison.ai.

The ruling impacts the medical technology sector, requiring health systems, investors, and lenders to track AI products and software based on regulatory clearance. While companies may possess advanced technology and strong clinical applications, their marketability hinges on FDA approval. Harrison.ai requested a partial exemption from the traditional 510(k) premarket-notification process for four radiology software categories, including tools to detect cancer lesions and analyze medical images.

The exemption would have applied only if a manufacturer had clearance for a related device. However, the FDA declined this proposal, emphasizing the necessity of premarket notification to ensure safety and effectiveness. Manufacturers must now submit 510(k) applications for clearance before marketing. The 510(k) process involves demonstrating that a device is substantially equivalent to one already on the market, which poses challenges for AI software that frequently updates.

This creates a financial concern for companies, as investors and lenders need to know which product version is cleared and how much growth relies on unapproved software. Health systems face a similar issue in linking clinical software to FDA-cleared versions, as they cannot assume newer versions retain the same approvals. Though the FDA retains flexibility for future less burdensome approaches, post-market monitoring remains a critical safety measure, supplementing rather than replacing the premarket gate.