I read with great interest the manuscript by Laohawetwanit et al, which evaluates the performance of Chat Generative Pre-trained Transformer (ChatGPT) in detecting and classifying colorectal adenomas from histopathological images.1 The study reports that ChatGPT achieved a median sensitivity of 74% and a specificity of 36% for adenoma detection, with an overall median accuracy of 56%. It further demonstrates variable accuracy in polyp classification, with notably low specificity and considerable inconsistency in diagnostic outputs, as indicated by kappa values ranging from 0.06 to 0.11. These findings highlight both the potential and limitations of applying advanced language models to the analysis of histopathological images in colorectal adenoma detection and classification.

The distinction between artificial intelligence (AI) …