Detection of retinal hemorrhage on head CT using deep learning: pediatric abusive-head-trauma AI workflow

The problem

Retinal hemorrhage is clinically important in evaluating possible abusive head trauma, but it's rarely identified on routine CT interpretation — the reference standard, dilated ophthalmologic exam, isn't always accessible. Head-CT AI is already crowded for intracranial hemorrhage, leaving this specific finding undetected on a scan already being acquired. And even a strong algorithm stays commercially inert without DICOM/PACS/EHR integration, AI governance, and a regulated deployment pathway — infrastructure no single academic team builds alone.

The solution

A patented AI workflow (U.S. Patent App. 18/421,718, Notice of Allowance received) that segments pediatric ocular globes from axial head CT, normalizes orientation, and applies deep learning to estimate retinal hemorrhage likelihood per eye — using CT data already acquired, no additional imaging required. UTRF is offering an exclusive or non-exclusive license to an imaging-AI platform, CT OEM, or enterprise-imaging company to pair this algorithm with existing infrastructure and carry it through reproducibility confirmation, multi-site validation, workflow definition, engineering, and regulatory strategy.

Benefits

• No new imaging or hardware — runs on CT scans already acquired
• Flags a clinically important finding radiologists often don't catch on routine CT
• Actionable output: triggers expedited ophthalmology review, not a diagnosis
• Fits directly into existing PACS, worklist, and enterprise triage platforms
• Differentiated indication in a crowded head-CT AI market
• Peer-reviewed foundation (JAMA Network Open) with defined performance metrics
• Clear, scoped path to product: validation, regulatory, and engineering needs already identified
• Useful where pediatric ophthalmology access is limited

Patent status

Patent Pending