The R.E.N.A.L. nephrometry scoring from CT reports with ChatGPT: example with proofs

Abstract

We read with great interest, curiosity the article by Toyama et al. published in the Japanese Journal of Radiology [1]. In this article, the authors compared the response of large language models (LLMs) to the questions of the Japan Radiology Board Examination, reported that ChatGPT-4 achieved the highest accuracy. Such studies provide an insight into the level of LLMs' knowledge of radiology and, as in this article, the literature is rapidly exploring the capabilities, limits of LLMs in various radiology felds. In this context, inspired by this article, we would like to present the performance of ChatGPT-4 on a diferent topic related to urogenital radiology