Clinical Decision Support
MS Risk Screening Agent
Spots early warning signs of multiple sclerosis in patient records and passes them to a specialist to judge.
01The problem
The early signs of multiple sclerosis are often small and spread across years of patient records: a visit about blurred vision here, numbness there. Each note on its own looks minor. Nobody has the time to read every record and connect them, so patients can wait years for a diagnosis.
02What it does
The system reads through patient records and flags people whose history shows an early risk pattern. Each flag comes with a plain explanation of why the patient was picked. A neurologist then decides what to do. The system finds candidates; it never diagnoses.
03How it works
Several agents share the work: one reads the records, one looks for the pattern, one writes the reasoning. The clinic can choose how much freedom the system has, from “only suggest” up to “prepare the referral for sign off”. Patient data is protected at every step.
04What came out of it
A working design for AI in a setting where mistakes are costly, built around one idea: the AI widens what a doctor can see, and the doctor still makes the call.
Built with
Want the technical detail?
The code and the full technical write up are on GitHub.