idw – Informationsdienst Wissenschaft
Nachrichten, Termine, Experten
Nachrichten, Termine, Experten
Medical reports written in technical terminology can pose challenges for patients. A team at the Technical University of Munich (TUM) has investigated how artificial intelligence can make CT findings easier to understand. In the study, reading time decreased, and patients rated the automatically simplified texts as more comprehensible and more helpful.
To simplify the original documents, the researchers used an open-source large language model operated in compliance with data protection regulations on the TUM University Hospital’s computers. An example: "The cardiomediastinal silhouette is midline. The cardiac chambers are normally opacified. [...] A small pericardial effusion is noted" was simplified by the AI as follows: "Heart: The report notes a small amount of fluid around your heart. This is a common finding, and your doctor will determine if it needs any attention."
Medicine needs to use understandable language
From the researchers’ perspective, making medical terminology accessible is more than a minor aid. "Ensuring that patients understand their reports, examinations, and treatments is a central pillar of modern medicine. This is the only way to guarantee informed consent and strengthen health literacy," says Felix Busch, assistant physician at the Institute for Diagnostic and Interventional Radiology and co-last author of the study, which was published in the journal "Radiology".
While previous research has shown that AI models can make specialist medical texts more comprehensible, little was known about their impact on actual patients. Therefore, the team included 200 patients who underwent CT imaging at the TUM University Hospital due to a cancer diagnosis. One half received the original report, while the other half received an automatically simplified version.
Reading time reduced, satisfaction high
The results were unambiguous: reading time fell from an average of seven minutes for the original reports to two minutes. Patients who received the simplified findings reported that they were much easier to read (81% compared with 17%) and easier to understand (80% compared with 9%). They also rated them as helpful (82% compared with 29%) and informative (82% compared with 27%) far more often. “Various objective measurements also confirmed the improved readability of the simplified reports,” says Felix Busch.
Future studies are needed to determine whether these advantages translate into measurable improvements in patient health outcomes. From the researchers’ perspective, however, the study clearly shows that patients can benefit from AI-supported simplification of medical reports by improving their understanding. "Providing automatically simplified reports as an additional service alongside the specialist report is conceivable. However, the prerequisite is the availability of optimized, secure AI solutions in the clinic," says Felix Busch.
Review by health professionals remains necessary
The team advises patients not to turn to a chatbot like ChatGPT as a stand-in doctor to simplify their report. “Aside from data protection concerns, language models always carry the risk of factual errors,” says Dr. Philipp Prucker, first author of the study. In the investigation, 6% of the AI-generated findings contained factual inaccuracies, 7% omitted information, and 3% added new information. Before the reports were provided to patients, however, they were reviewed for errors and corrected if necessary. “Language models are useful tools, but they are no substitute for medical staff. Without trained specialists verifying the findings, patients may, in the worst case, receive incorrect information about their illness,” Prucker concludes.
Additional Material for Media Outlets:
View this news item on tum.de: https://www.tum.de/en/news-and-events/all-news/press-releases/details/ai-helps-c...
TUM Corporate Communications Center contact:
Paul Hellmich
Media Relations
Tel. +49 (0) 89 289 22731
presse@tum.de
www.tum.de
Dr. Felix Busch
Technical University of Munich
TUM University Hospital
Institute for Diagnostic and Interventional Radiology
Phone +49 89 4140 1180
felix.busch@tum.de
Prucker et al. "A Prospective Controlled Trial of Large Language Model–based Simplification of Oncologic CT Reports for Patients with Cancer". Radiology (2025). DOI: 10.1148/radiol.251844.
Merkmale dieser Pressemitteilung:
Journalisten, Lehrer/Schüler, Studierende, Wirtschaftsvertreter, Wissenschaftler, jedermann
Ernährung / Gesundheit / Pflege, Informationstechnik, Medizin
überregional
Forschungsergebnisse, Wissenschaftliche Publikationen
Englisch
Sie können Suchbegriffe mit und, oder und / oder nicht verknüpfen, z. B. Philo nicht logie.
Verknüpfungen können Sie mit Klammern voneinander trennen, z. B. (Philo nicht logie) oder (Psycho und logie).
Zusammenhängende Worte werden als Wortgruppe gesucht, wenn Sie sie in Anführungsstriche setzen, z. B. „Bundesrepublik Deutschland“.
Die Erweiterte Suche können Sie auch nutzen, ohne Suchbegriffe einzugeben. Sie orientiert sich dann an den Kriterien, die Sie ausgewählt haben (z. B. nach dem Land oder dem Sachgebiet).
Haben Sie in einer Kategorie kein Kriterium ausgewählt, wird die gesamte Kategorie durchsucht (z.B. alle Sachgebiete oder alle Länder).
