OCR plus PII model redacts patient data burned into X-ray pixels on one GPU
MaziyarPanahi · x · 2026-10-29
A developer demonstrated a medical imaging de-identification pipeline: patient identifiers are burned into X-ray pixels, so clearing file headers isn't enough. An open OCR model reads every text label on the image, then a PII model blacks out identifiers — all on a single GPU.
The demo image shows 6 of 6 identifiers correctly redacted (name, ID, DOB, hospital, accession number, timestamp), with one false positive: the "Sin" side marker was mistakenly blacked out, showing such pipelines still need human review for non-identifier labels.
More from Safety
- a16z Podcast with Harvey's CLO: AI Expands Legal Services Rather Than Replacing Them — MattPerault · 2026-10-29
- UK budget allocates £40m for a National Agentic Security Unit, £120m more for AISI — davidmanheim · 2026-10-29
- Malware embeds nuke and bioweapon text to trip LLM safety refusals and dodge AI scanners — davidmanheim · 2026-10-29
- A new context poisoning attack tricks frontier LLMs into bioweapon content — davidmanheim · 2026-10-29
- MATS to Host AI Safety Talent Fair in Berkeley on November 19 — austinc3301 · 2026-10-29
- Deploying AI Agents Is Flying Blind — We Need Real-Time Monitoring — pstAsiatech · 2026-10-29