Realistic Super-Resolution Image Analysis for State of the Art 2D Contrast Enhanced Ultrasound Imaging

摘要

An ultrasound based super-resolution algorithm was developed for clinical 2D ultrasound imaging in contrast enhanced ultrasound (CEUS) mode. Synthetic and in vivo data from sheep ovaries were used. Development involved microbubble-specific detection and segmentation, resulting in optimal localisation, resulting in an image analysis algorithm that could be used with clinical data of a few minutes duration. Microscopy images from histology slices were available to allow comparison of the in vivo vascular network with that produced from the processed ultrasound data. A 5-fold resolution gain was demonstrated in vessel width, as vessels slightly under 120 μm were detected (λ = 257 μm), while the synthetic data demonstrated that the microbubble (MB) localization uncertainty is 30 μm. These results were achieved by processing video loops of under 2-minute duration. In conclusion, the above algorithm provided complete images of microvascular structure down to arteriole level from data that do not extend beyond a reasonable patient examination period.