Artificial intelligence is increasingly embedded within radiology workflows. In radiology, large language models may support reporting and communication tasks, while machine learning and deep learning models are increasingly used for image classification, segmentation, and quantitative analysis. This article is an educational review and practical guide rather than an experimental study. It summarizes how convolutional neural networks work, outlines the key stages of model development and validation, and discusses how modern language models may assist with code prototyping and troubleshooting. Although these tools may reduce technical barriers, clinically useful models still require appropriate datasets, external validation, transparent reporting, governance, and human oversight. This review aims to help radiologists understand the principles, potential, and limitations of convolutional neural networks in contemporary practice.
Thompson et al. (Fri,) studied this question.