System with confidence-based retroactive discrepancy flagging and methods for use therewith

This system helps doctors review medical scans by using artificial intelligence. It trains a computer vision model to spot abnormalities in scans, but uniquely, it uses existing medical reports to understand what's in the scans and how certain those reports are. This certainty, or 'confidence score,' helps the AI model become more reliable. When a new scan is processed, the system uses its trained AI to identify potential problems and displays this information to the doctor. The claims specify that the system uses a rules-based natural language processing approach to generate simple yes/no labels and continuous confidence scores from the medical reports.

Why it matters: Filed before the widespread adoption of powerful large language models (LLMs), the rules-based NLP system described in the claims for generating labels and confidence scores could now be significantly enhanced or replaced by more capable and adaptable LLM-driven approaches, simplifying data preparation.

Status: ExpiredProtection ended and was never renewed
How hard to build: SpecializedComputer vision, NLP, software integration

3 AI reinvention directions for this patent

See all three

AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.

Reinvent this with AI

Where this one came from

Every morning we go through millions of abandoned patents and pick out 15 worth your time. This was one of today’s.

This one

+10