Physical Ultrasound Modeling for Brain Applications (PUMBA)

Contact Persons

Walter Simson (walter.simson@tum.de)

Abstract

The aim of this cooperation project is the development of a robust deep learning-based ultrasonic simulator. Based on a large number of realistic ultrasound simulations, a parameterizable, generalized mapping between different tissue models is created, which models the physical relationships, especially of ultrasound, but which can be inverted more easily than analytical models. On the one hand, this facilitates the further development of algorithms for the detection of anatomical and pathological structures in ultrasound images, and on the other hand, areas of application are opened up in which ultrasound is currently not used due to its complexity. With the help of a bidirectional modality transfer between ultrasound and other imaging methods such as computed tomography or magnetic resonance tomography, a familiar visualization can be created for the attending physician. There is great clinical potential, also with regard to the potential radiation reduction, especially in the field of neurosurgery, which is why opening up this market is the primary goal of the project.

Keywords: Medical Imaging, Ultrasound, Simulation