Cloud Case Study: Real-Time Dental X-Ray QA
AWS ML Blog · rss · 2026-07-10
Henry Schein One used Amazon SageMaker AI to build a real-time dental X-ray quality verification system called Image Verify, shifting from post-event manual checks to instant feedback during capture.
Background and Impact
- In dental claims, up to 20% of insurance documents are initially rejected, often due to poor image quality.
- Since launch, the system quickly scaled to 10,000+ locations, processing over 11 million X-rays with a weekly increase of about 1.5 million.
- The company aims to expand globally to 40,000 locations.
How the System Works
- After image capture, the system identifies the X-ray type (e.g., bitewing, panoramic, periapical);
- Multiple specialized models evaluate sharpness, alignment, coverage, and completeness;
- These are aggregated into a 1–5 quality score returned to the front desk.
Key Architecture Optimizations
- SageMaker async inference: Scales based on queue depth rather than CPU utilization, ideal for GPU workloads;
- Instance optimization: Moving from ml.g6e.4xlarge to ml.g7e.4xlarge reduced median latency from 1.687s to 1.432s, and P90 from 2.45s to 2.196s;
- GPU count: Instance count dropped from 15 to 10, cutting GPU infrastructure by roughly 33%;
- Zero-downtime deployments and A/B testing ensure stable iterations;
- Multi-region deployment via AWS Cloud WAN covers the US, Europe, Canada, and Asia Pacific.
The article emphasizes that this solution assesses "whether the image is usable," rather than diagnosing conditions, making it much easier to scale within clinical workflows.
More from Infra
- A Firecracker-based platform says it can host 6,000 AI agents on one 256 GB server — maritime_sh · 2026-07-22
- oMLX 0.5.2 adds Mac menu-bar stats, low-bit decode kernels, and faster downloads — awnihannun · 2026-07-22
- Strangeworks launches Aura to turn enterprise ops into production optimization systems — whurley · 2026-07-22
- Graph workload 854.graph500 enters SPEC CPU 2026 as a new CPU benchmark — Prof_DavidBader · 2026-07-22
- Hybrid and local inference are emerging as a response to AI energy and token costs — dmitry140 · 2026-07-22
- NVIDIA details Vera CPU with 2x performance claims and a 22,000-core rack — ryanshrout · 2026-07-22