Truace
Subscribe

Patients preferred AI-simplified breast imaging reports in 63% of 2,965 ratings, a study finds

Study title: Large language model-based simplification of breast imaging reports: a prospective multicentre study with 2965 patient-based evaluations

A prospective multicentre survey study had 314 patients compare paired breast imaging report summaries, one written by a radiologist and one simplified by ChatGPT-4. Across 2,965 evaluations, patients preferred the AI version 63.3% of the time (1,876 evaluations), without knowing which report came from which source.

Read the study abstract

Objective To compare patient perceptions of radiologist-written and artificial intelligence (AI)-generated simplifications of breast imaging reports in terms of simplicity, comprehensibility, empathy, and overall preference. Materials and methods In this prospective multicentre survey-based study, 10 anonymized breast imaging reports (5 ultrasound, 5 mammography; one BI-RADS category 1-5 per modality) were simplified by ChatGPT-4 within a 50-word limit. After assessment by a panel of 20 physicians, paired radiol…

Truvace record2 min read1 sourceEvidence: Moderate · 50/100 — Survey · n = 314 · 1 independent sourcePeer-reviewedTRV-2026-2396 · cite or check it
Large language model-based simplification of breast imaging reports: a prospective multicentre study with 2965 patient-based evaluations"Medical laboratories and clinics in Kinwat" by Mouryan, CC BY-SA 4.0.

In brief

Patients gave high simplicity scores (4-5 on a 5-point scale) to 70.0% of AI reports against 43.4% of radiologist-written ones. For comprehensibility, the figures were 68.1% versus 40.4%. Patients increasingly read their own radiology reports, which are often hard to understand.

Comprehensibility was the strongest predictor of preference in both directions. Higher comprehensibility scores for the AI report raised the odds of choosing it (score 4: adjusted OR 1.87), while higher scores for the radiologist report lowered them (score 4: adjusted OR 0.45). The authors support supervised, complementary use of AI summaries.

Main points

  1. Ten anonymized breast imaging reports were used: five ultrasound and five mammography, covering BI-RADS categories 1 to 5 for each modality.
  2. ChatGPT-4 simplified each report within a 50-word limit, and a panel of 20 physicians assessed the pairs before patient evaluation.
  3. 314 patients produced 2,965 evaluations, rating simplicity, comprehensibility and empathy on a 5-point Likert scale and picking a preferred report.
  4. The origin of each report was concealed and each pair was shown in randomized order.
  5. AI reports were preferred in 63.3% of evaluations (1,876 of 2,965).
  6. High simplicity scores went to 70.0% of AI reports versus 43.4% of radiologist-written ones; high comprehensibility scores went to 68.1% versus 40.4%.
  7. In multivariable mixed-effects logistic regression, comprehensibility of the AI report (score 4: adjusted OR 1.87, 95% CI 1.08-3.23) increased odds of preferring AI.

The gain

In a multicentre survey of 314 patients, ChatGPT-4 simplified breast imaging reports were preferred in 63.3% of 2,965 evaluations and rated simple by 70.0% versus 43.4% for radiologist-written reports.

The full story

The study was a survey of patients rather than a clinical trial of outcomes. It tested how patients perceived simplicity, comprehensibility and empathy in short summaries of breast ultrasound and mammography reports.

Reports were chosen to span BI-RADS categories 1 to 5 for each of the two imaging types. ChatGPT-4 rewrote them in no more than 50 words, and 20 physicians reviewed the AI and radiologist pairs before patients saw them.

Patients rated each pair blind and in randomized order. Mixed-effects logistic regression looked for factors tied to choosing the AI report. Comprehensibility was the strongest predictor, with higher radiologist-report comprehensibility cutting the odds of choosing AI (score 4: adjusted OR 0.45, 95% CI 0.29-0.69).

The authors conclude that most patients preferred and found the AI versions clearer, and they back a complementary, supervised role for AI-generated summaries in clinical breast imaging.

Sources

  1. Peer-reviewedInsights into Imaging2026-11-02

The debate