Compressive Sensing for 3D/4D UltraSound Imaging: a Bayesian approach

  • Dooms, Ann (Administrative Promotor)
  • Bundervoet, Shaun, (Mandate)

Project Details

Description

Due to the steady and continuing advances in image acquisition technology an increasing number of applications can benefit from high resolution spatial and temporal data. However, a wide variety of domains, ranging from applications in the medical sector to the broadcasting industry, are faced with the problem of handling the growing amount of data at ever faster rates.
The objective of this project is to develop a generic approach to tackle to problem of high data rates at the receiver end of 2D+T architectures for the acquistion of 3D/4D ultrasound. We will pursue this goal by using a Bayesian learning methodology in combination with the compressive sensing paradigm.
In order to improve on current compressive sensing techniques using the information present in the incoming signals, we propose to use Bayesian modeling techniques. These techniques are typically very flexible and more easily adaptable to different use case then standard learning methodologies, the drawback being higher computational requirements.
The research will be divided into three parts, each corresponding to three essential components of the compressive sensing paradigm. The first is an investigation of existing sparsifying bases suitable for the acquisition methodoloy. This is crucial as a more concise representation of the incoming signals leads to a lower amount of samples needed in order to reconstruct this signal according to CS techniques. Therefore we will first investigate which bases are currently used in line array setups and which are already proposed for the 2D extension before incorporating these in a Bayesian learning strategy.
A second pillar in this project is the investigation of the sampling schemes used in different setups. There are some constrains on the amount and the location of the samples which can be left out. Usually this amounts to sampling the arriving signal uniformly at random. However a variety of pseudorandom schemes have been introduced depending on the application. Again learning approaches have been proposed in order to find sensing matrices which better suit a given dictionary. Moreover, this algorithm lends itself to a variety of adaptations, such as the incorporation of the Bayesian framework which we will do in this project.
The final component is the reconstruction itself. A large variety of iterative algorithms have been proposed for reconstructing the received signals/images from "incomplete" samples. These vary from general techniques to more customized versions like the Bayesian approach used for Ultrasound imaging in.
An important point of consideration which will be handled is the quality of the reconstruc- tion. The aforementioned mentioned Bayesian techniques all require significant calculations. Therefore the question arises whether or not the higher reconstruction quality also outweighs the increased complexity.
Combining the three components we obtain a flexible generic system which enables us to balance between restoration quality of the signals versus complexity and speed. Furthermore, due to the generalists approach our results can be readily extended to other acquisition technologies relying on the same principle.
AcronymIWT674
StatusFinished
Effective start/end date1/01/1431/12/17

Flemish discipline codes

  • Optics, electromagnetic theory
  • Materials science and engineering
  • Nanotechnology
  • Multimedia processing
  • Electronics
  • Applied mathematics in specific fields
  • Biological system engineering

Keywords

  • Low Power Cmos
  • Digital Image Processing
  • Numerical Linear Algebra
  • Embedded System Design
  • Image Reconstruction
  • Displays
  • Audio Processing
  • Light Detectors
  • Micro-Electronics Technology
  • Sige Bicmos Design
  • Satellite Image Analysis
  • Telemedicine
  • Medical Image Analysis
  • Inverse Problems
  • Video Compression
  • JPEGx
  • Neural Networks
  • Mine Detection
  • Vision
  • Digital Signal Processing
  • Electronic System Design
  • Machine Vision
  • Micro Electronics
  • Cmos Design
  • Chip Interconnects (Inter / Intra)
  • Humanitarian Demining
  • Speech Processing
  • Mpegx
  • Light Emitters
  • Pattern Recognition
  • Mm-Wave Technology
  • Robot Vision
  • Impedance Tomography
  • Image Compression
  • Light Modulators
  • Computer Aided Electronic Design
  • Medical Image Visualization
  • Motion Estimation And Tracking
  • Opto-Electronics
  • Multispectral Image Analysis
  • Electronics
  • Computer Vision
  • Image Processing
  • Industrial Visual Inspection
  • Image Analysis