Synthetic MR image generation of macrotrabecular-massive hepatocellular carcinoma using generative adversarial networks

• Published in: Diagnostic and interventional imaging 2023-05, Vol.104 (5), p.243-247
• Main Authors: Couteaux, Vincent, Zhang, Cheng, Mulé, Sébastien, Milot, Laurent, Valette, Pierre-Jean, Raynaud, Caroline, Vlachomitrou, Anna Sesilia, Ciofolo-Veit, Cybele, Lawrance, Littisha, Belkouchi, Younes, Vilgrain, Valérie, Lewin, Maité, Trillaud, Hervé, Hoeffel, Christine, Laurent, Valérie, Ammari, Samy, Morand, Eric, Faucoz, Orphee, Tenenhaus, Arthur, Talbot, Hugues, Luciani, Alain, Lassau, Nathalie, Lazarus, Carole
• Format: Article
• Language: English
• Subjects: Artificial intelligence; Carcinoma, Hepatocellular - diagnostic imaging; Carcinoma, Hepatocellular - pathology; Computer Science; Contrast Media; Data augmentation; Generative adversarial networks; Hepatocellular carcinoma; Humans; Liver Neoplasms - diagnostic imaging; Liver Neoplasms - pathology; Magnetic resonance imaging; Magnetic Resonance Imaging - methods; Medical Imaging
• ISSN: 2211-5684; 2211-5684
• DOI: 10.1016/j.diii.2023.01.003

•Generative adversarial network-based methods can be used to generate realistic synthetic MR images of rare liver tumors.•Generative adversarial network-based methods help generate a set of synthetic data mimicking the variability of real tumors in terms of size, shape and texture.•Generative adversarial network-based methods can leverage the high availability of routine MR images to generate a large-scale dataset of a rare liver tumor. The purpose of this study was to develop a method for generating synthetic MR images of macrotrabecular-massive hepatocellular carcinoma (MTM-HCC). A set of abdominal MR images including fat-saturated T1-weighted images obtained during the arterial and portal venous phases of enhancement and T2-weighted images of 91 patients with MTM-HCC, and another set of MR abdominal images from 67 other patients were used. Synthetic images were obtained using a 3-step pipeline that consisted in: (i), generating a synthetic MTM-HCC tumor on a neutral background; (ii), randomly selecting a background among the 67 patients and a position inside the liver; and (iii), merging the generated tumor in the background at the specified location. Synthetic images were qualitatively evaluated by three radiologists and quantitatively assessed using a mix of 1-nearest neighbor classifier metric and Fréchet inception distance. A set of 1000 triplets of synthetic MTM-HCC images with consistent contrasts were successfully generated. Evaluation of selected synthetic images by three radiologists showed that the method gave realistic, consistent and diversified images. Qualitative and quantitative evaluation led to an overall score of 0.64. This study shows the feasibility of generating realistic synthetic MR images with very few training data, by leveraging the wide availability of liver backgrounds. Further studies are needed to assess the added value of those synthetic images for automatic diagnosis of MTM-HCC.