What is a Finding Model?
A finding model describes a finding seen on an imaging exam, such as might be described by a radiologist or detected by an AI tool. The finding model specifies what the attributes of that finding are, and gives the finding itself and the attributes standardized names, so they can be referred to in a standard way—like the above.
Essentially, finding models define formats for structured descriptions of the results of radiological exams, enabling the interoperable, AI-driven medical care system of the future.
Why do we need finding models?
Finding models can empower large language models (LLMs) and other AI tools to bridge the gap between free text radiology reports into structured, actionable data. By bridging the gap between narrative content and standardized output, FMF accelerates clinical insight, research, and workflow integration for every stakeholder.
What is Finding Model Forge?
Finding Model Forge (FMF) empowers experts to create new finding models and improve existing ones.
FMF offers multiple workflows to combine the insight of experts with the power of artificial intelligence to create a complete system for representing all the details of the findings seen on imaging exams.
Workflows:
- Import existing information (ontologies, ACR/RSNA common data elements)
- Experts can update using text-based tools
- LLMs can use references to update (with expert review)
- LLMs can scan report text for examples to use to update (with expert review)
- Experts can review and indicate approval of a version of a finding model