AI-Generated Patient-Friendly MRI Fistula Summaries: A Pilot Randomised Study.

Area of Science:

• Gastroenterology and Radiology
• Artificial Intelligence in Medicine

Background:

• Perianal fistulising Crohn's disease (pfCD) impacts many Crohn's patients, necessitating frequent MRI monitoring.
• Standard radiology reports use technical language, hindering patient comprehension, increasing anxiety, and impeding shared decision-making.

Purpose of the Study:

• To evaluate the feasibility and safety of using AI (ChatGPT-4) to create patient-friendly MRI summaries for pfCD.
• To assess if AI summaries improve patient understanding, engagement, and shared decision-making compared to standard reports.

Main Methods:

• AI-generated patient-friendly summaries were created from de-identified MRI fistula reports.
• Clinicians assessed AI summaries for readability (Flesch-Kincaid) and hallucinations.
• Patients were randomized to review either AI summaries or original reports, rating them on readability, comprehensibility, utility, quality, and trustworthiness.

Main Results:

• Patients rated AI summaries significantly higher for readability, comprehensibility, utility, and quality, with fewer follow-up questions.
• Clinicians found AI summaries more readable and free of hallucinations.
• No clinically significant inaccuracies were identified in the AI-generated summaries.

Conclusions:

• AI-generated patient-friendly MRI summaries are feasible and safe for pfCD.
• These summaries show potential to enhance patient-clinician communication and streamline clinical workflows.
• Further research is needed to validate clinical utility and long-term acceptability.