A comparative analysis to assess anatomical illustrations via three AI-Driven Text-to-Image generators

In recent years the popularity of artificial intelligence (AI) has increased rapidly. While the machine itself can be built for deep learning in order to enhance its own efficiency through sophisticated networks, artificial intelligence (AI) uses machine-learning models to store, calculate, analyze, and even enhance extensive volumes of data that have to be retrieved whenever needed. Text-to-image AI models like Midjourney, Microsoft Bing Image Creator Powered by DALL- E, and Craiyon can generate artistic and impressive images. In this study, anatomical representations of the human ribs, brain, and lungs were produced by evaluating the above three AI-powered text-to-image producers. The generators were evaluated based on how well they represented the basic structure of ribs including correct number of ribs, false and floating ribs, sulcus and gyrus, as well as the structure of the cerebellum and thalamus. Not a single generator generated anatomically correct anatomical structures. It is required to add accurate images to the training databases to increase the accuracy. The study highlights the continued need of human medical illustrators, particularly in guaranteeing the availability of precise and understandable pictures. Improving their accuracy necessitates augmenting the training databases with anatomically correct images to help students.