AI drafts of heart-ultrasound reports matched doctors’ drafts in a Kaiser trial and cut cardiologists’ reading time by 46%

An artificial intelligence tool wrote first-draft reports on heart ultrasound exams that were about as accurate as drafts written by people, according to a Kaiser Permanente Northern California study presented Nov. 9 at the American Heart Association’s Scientific Sessions in Chicago.  With an AI draft in hand, cardiologists finished each report in 4.6 minutes on average, down from 8.5 minutes without help.

The heart ultrasound, called an echocardiogram, is the most common type of heart imaging, the association says.  A technician called a sonographer records moving pictures of the heart, and a cardiologist reviews them and writes a report on the heart’s structure and pumping.

How the trial worked

Researchers used 1,200 exams done over two weeks in March 2025 at 21 Kaiser medical centers in Northern California.  For each exam, the AI tool, a sonographer and a cardiologist each wrote a draft.  The drafts were then handed at random to cardiologists who did not know who wrote them.  The cardiologists made substantial changes to 50.6% of AI drafts and 49.7% of human drafts, a gap small enough to count as “noninferior,” or no worse.  They could not tell which drafts came from the AI.

“AI cannot, however, replace the human team,” said co-senior author Dr. Keane K. Lee, a cardiologist at Kaiser Permanente Santa Clara Homestead Medical Center.  The cardiologist still signs off on and is responsible for the final report, he said.

Limits

The results are preliminary.  Abstracts presented at the meeting are not peer-reviewed, the association notes.  The study did not test whether AI help improves patients’ health, and the cardiologists knew they were being studied, which may have changed how they worked.  The trial is listed on ClinicalTrials.gov, the federal registry of medical studies, as NCT07229300.

The U.S. could be short about 7,000 cardiologists by 2036, according to federal projections the association cites.  Lee said faster reads could mean quicker results for patients and more time with them for doctors.

Source:  American Heart Association news release, Nov. 9, 2026, Kaiser Permanente Division of Research study page and ClinicalTrials.gov record NCT07229300.