The aim of the project is to develop advanced techniques for efficient quantitative 3D microwave imaging. We intend to improve the current quality of 3D microwave reconstruction by joining our complementary expertise: electromagnetic forward modeling and inverse microwave reconstruction (INTEC), numerical aspects of nonlinear inverse problems (ETRO/IRIS) and statistical and multiresolution image modeling and image restoration (TELIN/IPI). Although the emphasis is on the algorithms and concepts, we will consider two applications: medical imaging for breast cancer detection (using phantoms only in this project) and non-destructive testing of reinforced concrete. For these applications, we can count on our scientific networks (e.g., Ghent University hospital and the department of civil construction) to provide feedback from the application domains. The speed of convergence of the algorithms and the quality of the reconstructed images strongly depend on the chosen cost criterion, which in turn depends on the prior models of the image content. Other important factors include the optimization strategy, the illumination and measuring configuration, the signal to noise ratio of the measuring chain and the complexity of the objects of interest. In the solution of inverse problems, often more than one formulation is available, with significantly different numerical properties, as it was shown for Electrical Impedance Tomography (EIT) / Electrical Capacitance Tomography (ECT). In the case of ill-posed inverse problems, application of optimized regularisation techniques are of utmost importance to stabilize solutions and recover numerical robust results. In many applications, careful interpretation of the underlying problem allows to reduce significantly the original dimensionality. Not only does this provide a reduction in the computational complexity, it often also allows to significantly improve upon the conditioning of the discredited problem. An example is 'shape based reconstruction' approach. We will research all of these aspects, with the ultimate aim of developing an algorithm that achieves an optimal compromise between image quality, processing time and illumination power. TELIN/IPI will focus on improving the regularization term of the cost function, through incorporation of the wavelet and MRF-based image models that have proven so successful in denoising and restoration. Also we will investigate the use of the linear sampling method to provide additional prior information (this could be useful in cases where the proposed technique is combined with other imaging modalities, rather than used on its own). Noise reduction (within the reconstruction process and/or through postprocessing) will also be an important focus of TELIN/IPI. INTEC will treat all electromagnetic aspects, such as the adaptation of the forward model in agreement with the considered applications, the development of qualitative imaging techniques to provide prior information and the definition of measurement configurations. ETRO/IRIS will address the numerical aspects associated with the actual reconstruction problem. As an alternative to the conventional SVD-based regularisation techniques, a new approach will be developed, building upon the interpolation properties of Radial Basis Functions (RBF's), whereby gradually a regularizing trajectory in the parameter space is unfolded.
|Effective start/end date||1/01/09 → 31/12/12|
- inverse problems
- microwave imaging
Flemish discipline codes
- Electrical and electronic engineering
- Mathematical sciences
- (Bio)medical engineering