The Technical University of Munich’s latest study unveils how AI assistance dramatically improves the readability and comprehension of CT scan reports for cancer patients, reducing reading time and increasing satisfaction rates.
Medical jargon can be a barrier for many patients trying to make sense of their diagnostic reports. To address this, a team from the Technical University of Munich (TUM) has harnessed the power of artificial intelligence to simplify CT findings, making them more accessible and understandable for cancer patients.
Study Highlights AI as a Game Changer
The TUM study, published in the journal Radiology, demonstrated significant improvements in patient comprehension and satisfaction.
Using an open-source large language model, the researchers simplified medical reports while adhering to data protection regulations on the TUM University Hospital’s secure servers.
A complex medical diagnosis such as, “The cardiomediastinal silhouette is midline. The cardiac chambers are normally opacified. […] A small pericardial effusion is noted,” was transformed by the AI into a more digestible version: “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.”
Impact on Patient Comprehension
Understanding medical terminology is crucial for patient care.
“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,” co-last author Felix Busch, an assistant physician at the Institute for Diagnostic and Interventional Radiology, said in a news release.
The research team included 200 cancer patients who had undergone CT imaging at TUM University Hospital. Half received their original reports, while the other half received simplified versions.
The results were notable: the reading time for the original reports averaged seven minutes, while the simplified texts took only two minutes to read.
Additionally, 81% of patients found the simplified reports easier to read versus 17% for original documents.
Similarly, 80% reported better understanding and 82% found the simplified versions helpful and informative compared to much lower percentages for the original texts.
Future Directions and Cautions
While the positive feedback is promising, further research is needed to determine if these improved comprehension levels directly translate to better health outcomes.
“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,” added Busch.
Despite the progress, the researchers advise against using AI without professional oversight.
“Aside from data protection concerns, language models always carry the risk of factual errors,” added first author Philipp Prucker.
In the study, some AI-generated reports contained inaccuracies (6%), omissions (7%) or additional information not present in the original reports (3%). Therefore, each AI-simplified report was reviewed and corrected by medical professionals before being given to patients.
“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 added.
Source: Technical University of Munich