Model-based image reconstruction for medical ultrasound

Most techniques that are used to reconstruct images from raw ultrasound signals are based on pre-defined geometrical processing. This type of image reconstruction typically has a low computational complexity and allows for real-time visualization. Since these techniques do not account for situation-specific parameters such as transducer characteristics and medium in-homogeneities, they cannot make proper use of the information that is contained in the raw ultrasound signals. In this paper, we explore the possibility of reconstructing images that best explain the measured ultrasound signals given the full ultrasound propagation model including all parameters. We build this model by measuring the spatiotemporal impulse response of the imaging transducer and, using the angular spectrum approach, estimate the ultrasound signal as it would originate from each individual image pixel position. An iterative search for the pixel combination that best explains the recorded signals provides the final image. We discuss the details of this model, provide experimental proof that this reconstruction allows for improved image quality, and extend our ideas to other imaging schemes such as compressive imaging.