Decoding radiology reports: Potential application of OpenAI ChatGPT to enhance patient understanding of diagnostic reports

被引：58

作者：

Li , Hanzhou ^{[1
]}

Moon, John T. ^{[1
]}

Iyer, Deepak ^{[1
]}

Balthazar, Patricia ^{[1
]}

Krupinski, Elizabeth A. ^{[1
]}

Bercu, Zachary L. ^{[1
]}

Newsome, Janice M. ^{[1
]}

Banerjee, Imon ^{[2
]}

Gichoya, Judy W. ^{[1
]}

Trivedi, Hari M. ^{[1
]}

机构：

[1] Emory Univ, Dept Radiol & Imaging Sci, Sch Med, 1364 Clifton Rd, Atlanta, GA 30322 USA

[2] Mayo Clin, Dept Radiol, Phoenix, AZ USA

来源：

CLINICAL IMAGING | 2023年 / 101卷

关键词：

Large language model; Patient-centered reports; Natural language processing; 21st century cures act; EXPERIENCE;

D O I：

10.1016/j.clinimag.2023.06.008

中图分类号：

R8 [特种医学]; R445 [影像诊断学];

学科分类号：

1002 ; 100207 ; 1009 ;

摘要：

Purpose: To evaluate the complexity of diagnostic radiology reports across major imaging modalities and the ability of ChatGPT (Early March 2023 Version, OpenAI, California, USA) to simplify these reports to the 8th grade reading level of the average U.S. adult.Methods: We randomly sampled 100 radiographs (XR), 100 ultrasound (US), 100 CT, and 100 MRI radiology reports from our institution's database dated between 2022 and 2023 (N = 400). These were processed by ChatGPT using the prompt "Explain this radiology report to a patient in layman's terms in second person: ". Mean report length, Flesch reading ease score (FRES), and Flesch-Kincaid reading level (FKRL) were calculated for each report and ChatGPT output. T-tests were used to determine significance.Results: Mean report length was 164 & PLUSMN; 117 words, FRES was 38.0 & PLUSMN; 11.8, and FKRL was 10.4 & PLUSMN; 1.9. FKRL was significantly higher for CT and MRI than for US and XR. Only 60/400 (15%) had a FKRL <8.5. The mean simplified ChatGPT output length was 103 & PLUSMN; 36 words, FRES was 83.5 & PLUSMN; 5.6, and FKRL was 5.8 & PLUSMN; 1.1. This reflects a mean decrease of 61 words (p < 0.01), increase in FRES of 45.5 (p < 0.01), and decrease in FKRL of 4.6 (p < 0.01). All simplified outputs had FKRL <8.5.Discussion: Our study demonstrates the effective use of ChatGPT when tasked with simplifying radiology reports to below the 8th grade reading level. We report significant improvements in FRES, FKRL, and word count, the last of which requires modality-specific context.

引用

页码：137 / 141

页数：5

共 27 条

[1]

American College of Radiology, 2021, ACR B

[2]

[Anonymous], 2021, BUSINESSWIRE

[3]

Ariyaratne S, 2023, SKELETAL RADIOL, V14

[4] Artificial hallucination: GPT on LSD? [J].

Beutel, Gernot ;

Geerits, Eline ;

Kielstein, Jan T. .

CRITICAL CARE, 2023, 27 (01)

[5] Patients' Use and Evaluation of an Online System to Annotate Radiology Reports with Lay Language Definitions [J].

Cook, Tessa S. ;

Oh, Seong Cheol ;

Kahn, Charles E., Jr. .

ACADEMIC RADIOLOGY, 2017, 24 (09) :1169-1174

[6]

Devlin J., 2018, NAACLHLT

[7] Readability of Patient Education Materials on the American Association for Surgery of Trauma Website [J].

Eltorai, Adam E. M. ;

Ghanian, Soha ;

Adams, Charles A., Jr. ;

Born, Christopher T. ;

Daniels, Alan H. .

ARCHIVES OF TRAUMA RESEARCH, 2014, 3 (02)

[8]

Flesch R, 2022, WRITE PLAIN ENGLISH

[9]

Friedman C., 1995, Natural Language Engineering, V1, P83, DOI 10.1017/S1351324900000061

[10] Automated Radiology Report Summarization Using an Open-Source Natural Language Processing Pipeline [J].

Goff, Daniel J. ;

Loehfelm, Thomas W. .

JOURNAL OF DIGITAL IMAGING, 2018, 31 (02) :185-192

← 1 2 3 →