I designed an AI workflow. Two years later, it became a filed patent.
A concept from an early-career design bootcamp that became a filed patent. An AI system that pre-reads scans, surfaces the most critical image in the worklist, and prioritises cases by severity — before the radiologist opens a single file.
Most people talk about AI as something coming. I've been designing with it as a material since 2020. This patent shows I approach it not as a feature to add, but as a system to reason about carefully.
The story
The problem
The worklist is a radiologist's entry point — a spreadsheet-like view of every assigned patient case. There's a priority indicator, but it's a single text column buried in rows. No images, no visual weight. The only way to know what a case actually needs is to open it.
At a 3–5% daily error rate across 100 cases, a radiologist could miss over 900 significant findings in a year. The problem wasn't skill or care. It was a system that made urgency easy to miss.
The solution
A machine learning algorithm pre-reads every scan in the background, surfaces the most critical image in the worklist, and ranks urgent cases higher — before the radiologist opens anything. Clicking through shows findings, confidence level, and EHR history.
After six weeks of the design bootcamp, we kept working. Two years later, the patent was filed.
.jpg)
