RadExtract: Radiology Report Structuring Demo

Powered by LangExtract + Gemini 2.5

article See the LangExtract Blog Post in Google for Developers open_in_new

warning This demonstration is for illustrative purposes only to show the baseline capabilities of LangExtract, the library that powers this demo. It does not represent a finished or approved product, is not intended to diagnose or suggest treatment for any disease or condition, and should not be used for medical advice.

License & Citation: If you use RadExtract or LangExtract in production or publications, please cite accordingly and acknowledge usage. Use is subject to the Apache 2.0 License. See for details.

LangExtract (LX) is a multi-purpose NLP extraction library that uses large language models such as Gemini to convert free-text into schema-controlled data. It learns from your few-shot examples (structured using LX's extraction schema) to identify and extract information, with every datum linked back to its exact words in the source.

Demo Overview

RadExtract uses LangExtract (LX) powered by Gemini 2.5 Flash or Gemini 2.5 Pro to convert radiology report findings into structured, optimized radiology reports with highlighted significant findings. By leveraging a prompt that describes the structuring task with LX's schema and a few select examples, LX processes free text into the structured output shown below. Leveraging Gemini's foundational knowledge, RadExtract can also process imaging modalities beyond those included in the prompt examples, such as the X-ray and ultrasound samples available below.

Interactive Features
Each extracted finding is directly grounded by LX (linked precisely back to its original words in the source text); hover over any structured item to see this exact textual origin highlighted. Clinical significance is visually highlighted: general findings are marked with yellow underlines, while significant findings have red underlines.

Clinical Background
Structured reporting helps ensure completeness, reduces ambiguity, and facilitates data sharing in radiology. For background on the value of structured radiology reports, see this European Society of Radiology paper.

Select a Report

1 Select a sample or paste your report

2 Click "Process" to start for pasted reports

3 Hover output findings to highlight source text

💡 Try tweaking a sample (remove sections, add extra findings, or paste your own report) to see how the demo responds.

💡 Tap any sample report to explore the structuring features. The keyboard stays closed so you can easily scroll and interact with highlighted findings.

Tip: Custom input is available only on desktop or laptop computers. If you're viewing this on a mobile device, please switch to a computer to enter your own reports.

Interface Controls

LX Generated Prompt • LX Structured Output • Use Cache expand_more

Input

LX Prompt

Output

LX Data

Model:

Use Cache LX Prompt LX Data

Findings: General Significant Grounding