From Canvas to Screen: Resurrecting Artists of the Past.

HomeRadiologyVol. 308, No. 1 PreviousNext Reviews and CommentaryFree AccessImages in RadiologyFrom Canvas to Screen: Resurrecting Artists of the PastRamin Javan , Navid MostaghniRamin Javan , Navid MostaghniAuthor AffiliationsFrom the Department of Radiology, George Washington University Hospital, 900 23rd St NW, Suite G2092, Washington, DC 20037 (R.J.); and California University of Science and Medicine School of Medicine, Colton, Calif (N.M.).Address correspondence to R.J. (email: [email protected]).Ramin Javan Navid MostaghniPublished Online:Jul 5 2023https://doi.org/10.1148/radiol.231118MoreSectionsPDF ToolsImage ViewerAdd to favoritesCiteTrack CitationsPermissionsReprints ShareShare onFacebookTwitterLinked In Supplemental material is available for this article.Generative AI art has made remarkable strides in recent years, demonstrating its prowess in creating images that closely resemble the styles of renowned artists from the past.—GPT-4 (1)Amidst the prodigious expanse of information, these ingenious algorithms have been most diligently tutored, thereby granting them the ability to apprehend the very quintessence of each masterful artist's unparalleled methodology and skill.—GPT-4 in the style of Charles Dickens (1)As these algorithms, akin to sorcery, do unfold, maturing and bettering with each fleeting hour, the prospect of their use in divers fields doth burgeon beyond measure, breaching the ramparts of invention and the bastions of creative thought, thereby transforming the very visage of medical imagery.—GPT-4 in the style of William Shakespeare (1)Integrating generative artificial intelligence (AI) art algorithms into medical imaging and radiology offers potential benefits for education, clinical practice, and research, such as creating pathology-specific images, enhancing quality, accelerating image acquisition, and reducing artifacts (2–5). However, this also raises cybersecurity concerns, including ransomware and fabricated pathologies. Prospective applications include cinematic rendering, multiplanar reconstruction, and three-dimensional segmentation. Additionally, AI-generated images can address the scarcity of high-quality, Health Insurance Portability and Accountability Act–compliant, annotated medical image data for research and model pretraining. Several text-to-image generative AI tools exist, including Midjourney (6), DALL-E 2, Stable Diffusion, and starryai. The images presented in Figures 1 and 2 were created in Midjourney through Discord using simple “/imagine” prompts of “An oil painting of the human [brain/heart] in the style of [artist name].” Please see Appendix S1 for instructions on how these images were generated. Of note, images generated by AI text-to-image tools are not protected by U.S. copyright law because they “are not the product of human authorship,” according to the nation's Copyright Office (7).Figure 1: The Heart as Envisioned by Artistic Legends. Images generated by Midjourney, version 5, show six representations of the human heart created in the distinctive styles of Salvador Dalí, Pablo Picasso, Leonardo da Vinci, Claude Monet, Vincent van Gogh, and Rembrandt (top left to bottom right). The diverse interpretations highlight the potential of generative artificial intelligence art in exploring the complexity of the heart, while staying true to the original artistic visions of these legendary painters.Figure 1:Download as PowerPointOpen in Image Viewer Figure 2: The Brain through the Eyes of the Masters. Images generated by Midjourney, version 5, show six representations of the human brain created in the distinctive styles of Salvador Dalí, Pablo Picasso, Leonardo da Vinci, Claude Monet, Vincent van Gogh, and Rembrandt (top left to bottom right). Each rendition offers a unique perspective on the intricacies of the brain and demonstrates the transformative power of generative artificial intelligence in capturing the essence of each artist's style.Figure 2:Download as PowerPointOpen in Image Viewer Disclosures of conflicts of interest: R.J. No relevant relationships. N.M. No relevant relationships.AcknowledgmentWe acknowledge parts of this article were generated with GPT-4 (powered by OpenAI's language model; http://openai.com) and Midjourney (https://www.midjourney.com/app/).References1. GPT-4. OpenAI. https://openai.com/gpt-4. Accessed May 1, 2023. Google Scholar2. Rudie JD, Gleason T, Barkovich MJ, et al. Clinical Assessment of Deep Learning-based Super-Resolution for 3D Volumetric Brain MRI. Radiol Artif Intell 2022;4(2):e210059. Link, Google Scholar3. Johnson PM, Recht MP, Knoll F. Improving the Speed of MRI with Artificial Intelligence. Semin Musculoskelet Radiol 2020;24(1):12–20. Crossref, Medline, Google Scholar4. Selles M, Slotman DJ, van Osch JAC, et al. Is AI the way forward for reducing metal artifacts in CT? Development of a generic deep learning-based method and initial evaluation in patients with sacroiliac joint implants. Eur J Radiol 2023;163:110844. Crossref, Medline, Google Scholar5. Cao C, Cui ZX, Liu S, Zheng H, Liang D, Zhu Y. High-Frequency Space Diffusion Models for Accelerated MRI. Electrical Engineering and Systems Science. arXiv preprint arXiv:2208.05481 [preprint]. https://arxiv.org/abs/2208.05481. Posted August 10, 2022. Accessed April 30, 2023. Google Scholar6. Midjourney, version 5. AI Art Generator App. https://www.midjourney.com/app/. Accessed April 28, 2023. Google Scholar7. Quach K. America: AI artwork is not authored by humans, so can't be protected by copyright. The Register. https://www.theregister.com/2023/02/24/copyright_ai_art_us/. Published February 24, 2023. Accessed April 30, 2023. Google ScholarArticle HistoryReceived: May 1 2023Revision requested: May 12 2023Revision received: May 13 2023Accepted: May 16 2023Published online: July 05 2023 FiguresReferencesRelatedDetailsRecommended Articles RSNA Education Exhibits RSNA Case Collection Vol. 308, No. 1 Supplemental MaterialMetrics Altmetric Score PDF download