Translation of atherosclerotic disease features onto healthy carotid ultrasound images using domain-to-domain translation

In this work, we evaluated a model for the translation of atherosclerotic disease features onto healthy carotid ultrasound images. An un-paired domain-to-domain – cycle Generative Adversarial Network (cycleGAN) was trained to translate between images arteries and pronounced disease. Translation performance using measurement wall thickness in original generated addition, explored different tissue segments (subcutaneous tissue, muscle, lumen, far wall, deep tissues), structural similarity index measure (SSIM) maps. Features were successfully translated (1.2 (0.33) mm vs 0.43 (0.07) mm, p < 0.001), while overall anatomy retained as SSIM value equal 0.78 (0.02). Exploration showed that both arterial subcutaneous tissues modified translation, but subject distortion some cases. The image quality influenced performance. results show can learn mapping diseased retaining anatomical contents. This is first study on atherosclerosis medical concept translating existing may serve purposes such education, cardiovascular risk communication health conversations, or personalized modelling precision medicine.