Compressive Sensing for Ultrasound Signal Reconstruction
Topic: Medical imaging is an exciting field where researchers are developing new techniques for diagnostic imaging and guidance of therapeutic interventions. How can compressive sensing reduce the required data rates and number of sensors for ultrasound imaging?Medical imaging is an exciting field where researchers are developing new techniques for diagnostic imaging and guidance of therapeutic interventions. Ultrasound imaging is one of these techniques that doctors use in their daily practice to get real-time images from the inside of the human body. These images normally show a thin 2D slice from a 3D object. For extending this to 3D imaging, special transducers with thousands of acoustic elements are needed. In a joint project between the Erasmus Medical Center and the TU Delft we are developing such a transducer. One of the challenges in this project is how to construct a 3D image from a limited number of measurements. Normally, the signals of all transducer elements are needed for this. We think that this challenge can be tackled with the use of new techniques that emerged from the exciting research-field of compressive sensing. This project will be a joint effort between the Thoraxcenter Biomedical Engineering group of the Erasmus Medical Center in Rotterdam and the Circuits and Systems group of the Delft University of Technology.
AssignmentIn this master project you will try to solve the problem of reconstructing a full 3D image or 3D acoustic wavefield using only a limited number of combined element signals. In the figure, a simplified schematic shows how these compressive signals are acquired. The squares (blue and white) in this figure depict individual acoustic elements within a larger array. Because we only have a limited number of analog-to-digital converters (ADC) we cannot acquire all the element signals individually. Instead, the signals received by the blue elements will be column-wise summed and fed into one ADC. In the next acquisition event we will change which signals will be summed and which not (white elements). We expect from compressive sensing theory that by using a limited number of these acquisition events under appropriate constraints we are able to reconstruct the full acoustic field equally well as from all the element signals individually. In the first part of the project you will get acquainted with the mathematical models and simulation tools that describe ultrasound signals and image reconstruction. With this in hand you will perform experiments with a conventional linear array transducer where groups of elements will be summed into one signal, resulting in a limited number of compressed signals. Your task will then be to reconstruct the complete image from this limited number of signals. In the end we hope that we can make 3D images using only a few signals and of course your technique!
RequirementsFor this project, we are looking for a master student in either electrical engineering, applied physics, mathematics, computational science or any related study. Furthermore we are looking for a student who is enthusiastic about working in a medical research setting and is committed to make this project a success. Programming skills in Matlab and/or C/C++ and basic knowledge of signal processing are highly appreciated. Good spoken and written English is a must. Expected project duration is about 9 months.
prof.dr.ir. Geert Leus
Circuits and Systems Group
Department of Microelectronics
Last modified: 2015-04-30