Numerical Simulation of Intravascular Ultrasound Images Based on Patient-Specific Computed Tomography

Abstract

Intravascular ultrasound (IVUS) provides detailed imaging of the artery circumference. Over the past years, the interest in artificial intelligence (AI) for interpretation and automatic analysis of IVUS images has grown. Development of such algorithms typically requires considerable amounts of annotated data. However, manual annotation of IVUS data is time-consuming and expensive. An alternative solution would be the simulation of IVUS data, which yields images with all necessary ground-truth data available. Therefore, in this study, we present an IVUS simulator to simulate realistic IVUS data based on computed tomography (CT) images. The IVUS transducer is modeled accurately, which is reflected in the in vitro and in silico measurements of the point-spread function (PSF) and speckle size. The capability of simulating realistic IVUS images is showcased on an in vivo co-registered CT-IVUS dataset of two patients with an abdominal aortic aneurysm (AAA). Quantitative results, expressed in terms of the Jensen-Shannon divergence (JSD), speckle signal-to-noise ratio (sSNR), and contrast-to-noise ratio (CNR), reveal the high similarity between the in vivo and in silico IVUS images. The proposed simulator is promising for ultrasound data generation, enabling the generation of IVUS images with the desired ground truth.