Скачать или смотреть Michael Mayerhofer - AI for ultrasound imaging from a probe with an ergodic relay

Ultrasound imaging is one of the most used clinical imaging methods, thanks to its portability, non-invasiveness, and real-time capabilities. In my research, I developed a neural network capable of transforming radiofrequencies captured by an ultrasound probe, with an ergodic relay (prism), into discernable images. Unlike conventional methods that need data collected by multiple elements, using a prism significantly lowers the cost of making probes since only one element is needed. However, this efficiency leads to longer computation times. By training the neural network, I addressed the increased computational demand, making it possible to quickly solve the inverse problem. This improvement promises faster image creation in ultrasound imaging, offering a more affordable option for practitioners, despite the trade-off of longer computation times.