GPT IDs Errors in Reports With Same Accuracy as Radiologists

TUESDAY, April 30, 2024 (HealthDay News) -- GPT-4 detects errors in radiology reports with the same accuracy as that of radiologists, according to a study published online April 16 in Radiology.

Roman Johannes Gertz, M.D., from University of Cologne in Germany, and colleagues assessed the effectiveness of GPT-4 in identifying common errors in radiology reports, focusing on performance, time, and cost-efficiency. The analysis included 200 radiology reports (radiography and cross-sectional imaging [computed tomography and magnetic resonance imaging]) compiled between June 2023 and December 2023 at one institution, with 150 errors from five common error categories (omission, insertion, spelling, side confusion, and other) intentionally inserted into 100 of the reports. Results were compared from GPT-4 and six radiologists (two senior radiologists, two attending physicians, and two residents).

The researchers found that GPT-4 (detection rate, 82.7 percent) matched the average detection performance of radiologists independent of their experience (senior radiologists, 89.3 percent; attending physicians, 80.0 percent; residents, 80.0 percent). Only one senior radiologist outperformed GPT-4 (detection rate, 94.7 percent). Compared with the fastest human reader in the study, GPT-4 required less processing time per radiology report (mean reading time, 25.1 versus 3.5 seconds). Further, the use of GPT-4 resulted in lower mean correction cost per report than the most cost-efficient radiologist ($0.03 versus $0.42).

"The study addresses critical health care challenges such as the increasing demand for radiology services and the pressure to reduce operational costs," Gertz said in a statement. "Ultimately, our research provides a concrete example of how artificial intelligence, specifically through applications like GPT-4, can revolutionize health care by boosting efficiency, minimizing errors and ensuring broader access to reliable, affordable diagnostic services -- fundamental steps toward improving patient care outcomes."

Several authors disclosed ties to the pharmaceutical industry.

Abstract/Full Text