Year: 2024
Type: article
Abstract: Use of ChatGPT to Assign BI-RADS Assessment Categories to Breast Imaging ReportsLarge language models (LLMs) have shown early promise in radiology applications, including answering patient questions about breast cancer screening [1], providing decision support for breast imaging appropriateness [2], and streamlining structured reporting [3].LLMs such as ChatGPT GPT-4 (March 2023 release; OpenAI) could potentially improve breast imaging workflow by automating BI-RADS assessment category assignment during report creation.This study's aim was to assess the utility of ChatGPT in automating the assignment of BI-RADS assessment categories based on the text of breast imaging reports.This retrospective study received an institutional review board exemption.For study purposes, a fellowship-trained breast imaging radiologist (M.B., 8 years of posttraining experience) manually created 250 factitious breast imaging reports, including 50 reports for screening mammography examinations (25 with BI-RADS category of 0; 25 with BI-RADS category of 1 or 2), 100 reports for diagnostic mammography and/or ultrasound examinations (BI-RADS category ranging from 1 to 6), and 100 reports for MRI examinations (for high-risk screening or diagnostic evaluation; BI-RADS category ranging from 1 to 6) [4].The factitious a March 2023 release of GPT-4 (OpenAI).b Represents fraction of factitious reports for which GPT-4 provided same BI-RADS category for all three tests, when GPT-4 was provided the same prompt and same report in three independent tests.c Radiologist-assigned reference standard of BI-RADS category 0 in 25 reports and category 1 or 2 in 25 reports.dRadiologist-assigned reference standard of BI-RADS category 1 or 2 in 30 reports, category 3 in 30 reports, category 4 or 5 in 30 reports, and category 6 in 10 reports.e Radiologist-assigned reference standard of BI-RADS category 1 or 2 in 25 reports, category 3 in 25 reports, category 4 or 5 in 25 reports, and category 6 in 25 reports. ARRS is accredited by the Accreditation Council for Continuing Medical Education (ACCME) to provide continuing medical education activities for physicians.The ARRS designates this journal-based CME activity for a maximum of 1.00 AMA PRA Category 1 Credit™.Physicians should claim only the credit commensurate with the extent of their participation in the activity.Use the "Claim Credit" link to access the CME activity.Downloaded from ajronline.orgby 198.1.
DOI: 10.2214/ajr.24.31093
Authors: Hana L. Haver, Paul H. Yi, Jean Jeudy, Manisha Bahl
Language: English
Subfield: Health Informatics
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W4396722336