×

A deconvolution framework with applications in medical and biological imaging. (English) Zbl 1304.92031

Heidelberg: Univ. Heidelberg, Naturwissenschaftlich-Mathematische Gesamtfakultät (Diss.). xl, 197 p. (2012).
Summary: A deconvolution framework is presented in this thesis and applied to several problems in medical and biological imaging. The framework is designed to contain state of the art deconvolution methods, to be easily expandable and to combine different components arbitrarily. Deconvolution is an inverse problem and in order to cope with its ill-posed nature, suitable regularization techniques and additional restrictions are required.
A main objective of deconvolution methods is to restore degraded images acquired by fluorescence microscopy which has become an important tool in biological and medical sciences. Fluorescence microscopy images are degraded by out-of-focus blurring and noise and the deconvolution algorithms to restore these images are usually called deblurring methods. Many deblurring methods were proposed to restore these images in the last decade which are part of the deconvolution framework. In addition, existing deblurring techniques are improved and new components for the deconvolution framework are developed. A considerable improvement could be obtained by combining a state of the art regularization technique with an additional non-negativity constraint. A real biological screen analysing a specific protein in human cells is presented and shows the need to analyse structural information of fluorescence images. Such an analysis requires a good image quality which is the aim of the deblurring methods if the required image quality is not given.
For a reliable understanding of cells and cellular processes, high resolution 3D images of the investigated cells are necessary. However, the ability of fluorescence microscopes to image a cell in 3D is limited since the resolution along the optical axis is by a factor of three worse than the transversal resolution. Standard microscopy image deblurring techniques are able to improve the resolution but the problem of a lower resolution in direction along the optical axis remains. It is however possible to overcome this problem using Axial Tomography providing tilted views of the object by rotating it under the microscope. The rotated images contain additional information about the objects which can be used to improve the resolution along the optical axis. In this thesis, a sophisticated method to reconstruct a high resolution axial tomography image on basis of the developed deblurring methods is presented.
The deconvolution methods are also used to reconstruct the dose distribution in proton therapy on basis of measured PET images. Positron emitters are activated by proton beams but a PET image is not directly proportional to the delivered radiation dose distribution. A PET signal can be predicted by a convolution of the planned dose with specific filter functions. In this thesis, a dose reconstruction method based on PET images which reverses the convolution approach is presented and the potential to reconstruct the actually delivered dose distribution from measured PET images is investigated.
Last but not least, a new denoising method using higher-order statistic information of a given Gaussian noise signal is presented and compared to state of the art denoising methods.

MSC:

92-02 Research exposition (monographs, survey articles) pertaining to biology
92C55 Biomedical imaging and signal processing
94A08 Image processing (compression, reconstruction, etc.) in information and communication theory
68U10 Computing methodologies for image processing
PDFBibTeX XMLCite
Full Text: Link