How the United States Deploys AI in Medical Imaging: FDA-Cleared Devices and Clinical Practice

The United States is the world's largest market for AI-enabled medical imaging software. As of 2025, the U.S. Food and Drug Administration (FDA) has cleared close to a thousand AI/ML-enabled medical devices, the majority of which are imaging-related. Understanding how American hospitals deploy these tools — and how regulators approve them — offers a useful reference for the adoption of imaging AI in China.

How FDA approves imaging AI

Most imaging AI software enters the U.S. market through the FDA's 510(k) pathway, where a new device is shown to be substantially equivalent to an already cleared device. A smaller number go through the more rigorous De Novo pathway for genuinely novel products.

Well-known examples include:

• IDx-DR (2018): the first FDA-authorized autonomous AI system, which detects diabetic retinopathy from retinal images without a specialist interpreting the scan;
• Viz.ai: AI that analyzes CT scans to detect large vessel occlusion in stroke and automatically notifies the care team;
• AI for lung nodules and breast imaging: dozens of cleared systems assist radiologists in detecting pulmonary nodules, fractures and suspicious breast findings.

How American hospitals use it in practice

Deployment in the U.S. is increasingly "radiologist-centric": AI findings are embedded directly into radiology worklists, often through enterprise AI orchestration platforms that route studies to the most appropriate algorithm. Common patterns include triage tools (flagging critical cases first), detection tools (highlighting suspicious regions) and quantification tools (measuring volumes or density).

Reimbursement remains a key driver. New CPT codes for AI-assisted image reading were introduced in 2022, and more codes have followed, giving hospitals a financial incentive to adopt AI that improves workflow efficiency.

Challenges

U.S. experience also highlights pitfalls: integration with existing PACS/RIS systems can be slow; alert fatigue occurs when algorithms produce too many false positives; and algorithms often need site-specific validation because performance can vary across populations, devices and protocols.

Lessons for China

For China, the U.S. experience confirms that regulatory clarity (NMPA registration), enterprise integration (HIS/PACS), payment mechanisms and real-world validation are the four pillars of sustainable imaging-AI adoption. It also shows that the most successful products solve a specific clinical bottleneck rather than trying to be everything at once.

At yoyolink, we apply the same discipline to GI endoscopy: one clear clinical problem — supporting early-cancer detection and quality control in gastroscopy and colonoscopy — solved with dedicated imaging models that are transparent, traceable and free from AI hallucination.

• Medical Imaging AI Whitepaper — the full technology and policy review
• yoyolink product overview — how dedicated endoscopy AI works
• Contact us — request materials or a demo

Triage, reporting and secure integration

US radiology increasingly relies on triage tools to prioritize critical cases. yoyolink applies the same logic to endoscopy: patent CN112233752A/B supports criticality assessment, report print priority and neural-network physician matching, so high-risk examinations move ahead of the queue. Integration and security are supported by patent CN112714233B (block-based encrypted image transmission) together with HIS/PACS interfaces and hospital-local storage.

Patents & IP

Basemed Medical protects its technology with granted invention patents:

• "Information Processing Method and Device for Endoscopic Examination Report" (CN112233752A/B, co-filed with Zhejiang University): intelligently processes endoscopy reports — detects lesions, assesses criticality, prioritizes report printing, and matches patients with suitable physicians via neural networks. View | PDF
• "Method and System for Intelligent Transmission of Endoscopic Images Based on Block Decoding" (CN112714233B): transmits endoscopic images in blocks with multi-layer key information and quality detection, improving transmission quality, security and efficiency. View | PDF

See the full portfolio: About · Patents & IP