Mayo Clinic Tool Summarizes Patient Histories for Radiologists

A multi-agent AI system called GRACIE condenses fragmented electronic health records into structured summaries at the point of image interpretation.

Updated on Nov. 19, 2026 in Artificial Intelligence

Mayo Clinic Tool Summarizes Patient Histories for Radiologists

Live Poll

Would you trust a doctor whose diagnosis was partly informed by AI summaries?

Mayo Clinic researchers presented GRACIE, a large language model-based multi-agent system that summarizes patient clinical histories for radiologists. In a survey conducted in March and April 2026, nearly 93% of respondents actively used the tool.

Why it matters

Synthesizing complex clinical histories from fragmented health records is time-consuming and adds to radiologists' cognitive burden. Automated, citation-enabled summarization may let clinicians focus more on primary image analysis.

GRACIE synthesizes a median of 109 EHR entries per patient, including lab results, clinical notes, and prior radiology reports, into a time-ordered summary. Users reported median time savings of five minutes per case, but the survey relied on self-reported estimates, and its effect on diagnostic accuracy was not reported.

The players

Mayo Clinic

A nonprofit academic medical center based in Rochester, Minnesota, known for its integrated radiology and clinical informatics programs.

The details

GRACIE presents a structured, time-ordered patient summary at the time of interpretation. About three-quarters of surveyed users said it uncovers facts they might otherwise have missed, and nearly all said it saves time and reduces the burden of searching charts.

Timeline

  1. March and April 2026: Survey of radiologists, fellows, and residents conducted.

  2. Tuesday: GRACIE presented in a scientific session.

The Tech Race

GRACIE extends Mayo Clinic's clinical informatics work on electronic health record integration by applying large language models to summarization. It positions health systems to use LLM-based tools as an alternative to the manual chart-review workflows that radiology has long relied on.

If adopted in clinical settings, tools like GRACIE could shorten the time radiologists spend reviewing patient charts before interpreting images. Patients would not see the tool directly, but faster and more complete chart review could affect how quickly imaging results are reported.

The takeaway

Clinical AI tools are increasingly aimed at the unglamorous work of assembling records, not just interpreting images. Evaluating them requires measuring accuracy and outcomes, not only user satisfaction.

Further reading

For broader context on how these systems are reshaping medicine, explore more Artificial Intelligence coverage.

Live Poll

Would you trust a doctor whose diagnosis was partly informed by AI summaries?