Moritz Blume Technische Universität München Boltzmannstr. 3 85748 Garching bei München Germany OR Institute of Particle Physics - IFIC Edificio Institutos de Investigacion Apartado de Correos 22085 E-46071 Valencia Spain

Research Topics

My general research interests are image restoration and reconstruction in medical imaging. Especially, I am interested in multi-view deconvolution and image reconstruction for positron-emission-tomography (PET).

Positron-Emission-Tomography Motion Compensation

Positron-emission-tomography (PET) is nowadays a well-established tool for diagnosing, staging and monitoring diseases in patients. However, due to the long acquisition times which are necessary for a PET scan, patient movements are often unavoidable. Patient motion during data acquisition can cause image blurring and severe artifacts in the reconstructed image, thus posing a significant problem for further diagnosis and treatment.

Reconstruction algorithms that incorporate patient motion are still at a premature level. My interest is to improve known methods, e. g. by incorporating partial motion information.

Multi-View Deconvolution

"Wer misst misst Mist" - direct translation into english (trying to conserve the alliteration...): "Whoever measures measures muck". Free translation into english: "measurements are error-prone". Whoever tried to measure a physical quantity knows about the truth of this phrase. Measuring devices that everyone of us knows are digital cameras, microscopes, ultrasound... however, even if the images taken by a digital camera look quite good, the introductory phrase holds: measurements are error-prone.

I am looking at blurring. In many biomedical applications, blurring depends on the point of view. For example, two observations from two different points of views of a 3D microscope could look like this:

Each observation offers additional information about the original image. Already on the base of these two observations, my algorithm yields the following restored image:

This restoration is done without any knowledge about the degradation process. Such methods are also called blind deconvolution methods. Even more realistic scenarios - like shift-variant blurring - can be tackled with our method. For more details, please refer to the literature listed below.

Publications

2010
	M. Blume, A. Martinez-Möller, A. Keil, N. Navab, M. Rafecas Joint Reconstruction of Image and Motion in Gated Positron Emission Tomography IEEE Trans. Med. Imag., vol. 29, no. 11, pp. 1892-1906, November 2010 (bib)
	M. Blume, A. Keil, N. Navab, M. Rafecas Joint Reconstruction of Image and Motion for PET: Displacement Fields Versus a B-Spline Motion Model Proc. IEEE Nuclear Science Symposium and Medical Imaging Conference, Knoxville, TN, U.S.A., October 2010 (bib)
	M. Blume, M. Rafecas Joint Reconstruction of Image and Motion for PET Using Linear Diffusion Regularization Congreso Anual de la Sociedad Espa\~nola de Ingenier\'ia Biom\'edica, Madrid, Spain, November 2010 (bib)
2009
	M. Blume, A. Keil, N. Navab, M. Rafecas Blind Motion Compensation for Positron-Emission-Tomography SPIE Medical Imaging, Orlando, FL, U.S.A., Feburary 2009 (bib)
	M. Blume, A. Martinez-Möller, A. Keil, N. Navab, M. Rafecas Simultaneous Reconstruction of Image and Motion in Gated Positron-Emission-Tomography Proc. IEEE Nuclear Science Symposium and Medical Imaging Conference, Orlando, FL, U.S.A., October 2009 (bib)
	M. Blume, A. Martinez-Möller, A. Keil, N. Navab, M. Rafecas Joint Reconstruction of Image and Motion in Gated Positron-Emission-Tomography 10th International Meeting on Fully Three-Dimensional Image Reconstruction in Radiology and Nuclear Medicine, Beijing, China, September 2009 (bib)
	T. Felder, M. Blume, J. Oliver, M. Rafecas ML-EM Implementation on a GPU: Avoiding Simultaneous Read-Modify-Write Processes 10th International Meeting on Fully Three-Dimensional Image Reconstruction in Radiology and Nuclear Medicine, Beijing, China, September 2009 (bib)
2008
	M. Blume, M. Rafecas, S. I. Ziegler, N. Navab Combined Motion Compensation and Reconstruction for PET Proc. IEEE Nuclear Science Symposium and Medical Imaging Conference, Dresden, Germany, October 2008 (bib)
2007
	W. Wein, M. Blume, U. Leischner, H.-U. Dodt, N. Navab Quality-based Registration and Reconstruction of Optical Tomography Volumes Medical Image Computing and Computer-Assisted Intervention (MICCAI 2007), Brisbane, Australia, October 2007 (bib)
	M. Blume, D. Zikic, W. Wein, N. Navab A New and General Method for Blind Shift-Variant Deconvolution of Biomedical Images Medical Image Computing and Computer-Assisted Intervention (MICCAI 2007), Brisbane, Australia, October 2007 (bib)

Masters Thesis (Diplomarbeit)

M. Blume, D. Zikic, W. Wein, N. Navab Multi-View Deconvolution of Biomedical Images Technische Universität München, 2007

Reports and Tutorials

	M. Blume, W. Wein, U. Leischner, H. U. Dodt, N. Navab Image Registration and Merging for a new Optical Tomography Technische Universität München, 2005.
	M. Blume Expectation Maximazation: A Gentle Introduction Technische Universität München, 2008.
	M. Blume Expectation Maximization for Emission Tomography Technische Universität München, 2008.
	M. Blume An Introduction to Probability Theory Technische Universität München, 2008.

My Favorite Links