Meeting Banner
Abstract #1991

Evaluating Large Language Model’s Potential to Optimize Structured MRI Reporting in Breast Cancer Diagnosis

Yang Song1, Muzhen He2, Ailing Wang3, Chenglong Wang3, and Guang Yang3
1MR Research Collaboration Team, Siemens Healthineers Ltd, Shanghai, China, 2Department of Radiology, Shengli Clinical Medical College of Fujian Medical University, Fujian Provincial Hospital, Fuzhou University Affiliated Provincial Hospital, Fuzhou, China, 3Shanghai Key Laboratory of Magnetic Resonance, East China Normal University, Shanghai, China

Synopsis

Keywords: Breast, Diagnosis/Prediction

Motivation: Motivated by the need for precise breast cancer diagnosis, this study investigates whether LLMs, like ChatGPT, can optimize MRI reporting in line with BI-RADS guidelines.

Goal(s): We aimed to convert clinical reports into structured data and evaluate the feasibility for clinical application.

Approach: Using OpenAI's text-davinci-003 model on reports from 237 patients, we found Kappa values indicated variable agreement.

Results: The high sensitivity of the model suggests effective capture of positive features. ROC analysis showed that GPT's diagnostic performance was comparable to physician-annotated tests, particularly for HR and HER2.

Impact: This study highlights the promising role of AI in radiology, potentially enhancing diagnostic accuracy and supplementing radiological expertise, especially in multilingual and resource-limited settings.

How to access this content:

For one year after publication, abstracts and videos are only open to registrants of this annual meeting. Registrants should use their existing login information. Non-registrant access can be purchased via the ISMRM E-Library.

After one year, current ISMRM & ISMRT members get free access to both the abstracts and videos. Non-members and non-registrants must purchase access via the ISMRM E-Library.

After two years, the meeting proceedings (abstracts) are opened to the public and require no login information. Videos remain behind password for access by members, registrants and E-Library customers.

Click here for more information on becoming a member.

Keywords