Medical Image Processing

Department of Cybernetics, Karlovo náměstí 13
http://cmp.felk.cvut.cz/~kybic

About us

doc. Dr. Ing. Jan Kybic, group leader


Other members of the team.

Our research

We develop new algorithms for biomedical image processing. We process images from different modalities, such as magnetic resonance, ultrasound, computed tomography, or microscopy. We work in 2D, 3D and 4D. We know how to preprocess the data, how to register, segment, model, reconstruct and classify them. We use techniques from image processing, numerical mathematics, as well as machine learning.

Motivation

The amount of clinically produced image data increases every day. Many classical modalities (such as X-ray scanners) now provide digital data, 3D modalities (such as MRI - Magnetic Resonance imaging - or CT - Computed tomography) are now used on daily basis in clinical practice and their resolution increases, increasing also the amount of the produced data. The bottleneck of the diagnosis is no longer the acquisition technique but the availability of radiology experts. Many american hospitals are now routinely sending their data to China or India for evaluation.

We intend to develop tools to facilitate and speed up the radiologist's job, for example by emphasizing the changes since the previous exam or pointing out areas of potential interest. This paves a way to fully automatic diagnosis systems of the future...

In biology the amount of the generated data has also seen an enormous growth - 3D microscopy is becoming common, resolution is increasing, and robotized laboratories can automatically prepare and acquire a large number of samples. However, it is impossible for a human to examine for example all cells in a sample to see if they are infected by a parasite or if they contain a genetic anomally. A human takes months to make a drawing of nervous fibers in a piece of tissue smaller than a milimeter. There is therefore a great potential for computer algorithms capable of speeding up the analysis of the acquired data. The advantage is that, unlike in medicine, a small percentage of errors can usually be tolerated.

Research projects

  • Algorithms for tracking of neuronal fibers (processes) in electron microscopy (especially 3D, FIB) and light microscopy (bright field and confocal) images, extracting the neuronal trees, and alignment of electron microscopy and optical microscopy volumes.
  • Algorithms for image registration, i.e. finding a geometrical transformation between two images of the same or similar objects. Registration is used to detect and quantify movement or changes, and to align images taken at different times or using different imaging devices.
  • Developing new image similarity criteria and methods for uncertainty estimation for image registration.
  • Fast and robust detection and localization of surgical tools such as needles or electrodes from 3D ultrasound images for visualization and automatic guidance purposes.
  • Numerical methods for ultrasound elastography using standard medical ultrasound scanner. Elastography can determine mechanical properties of tissues, which can help for example to detect tumors.
  • Detection of lung nodules from CT images to prevent lung cancer.
  • Algorithms for parallel MRI reconstruction, used for faster acquisition.
  • Spatial reconstruction of brain activity from EEG and MEG.
  • Respiratory movement modeling for movement compensation in radiation therapy.
  • Fast algorithms for 3D segmentation of medical images.
  • Malaria diagnosis from microscopy images of blood samples.
  • Quantification of opacity in cardiological angiography for heart perfusion evaluation.
  • Automatic quantitative analysis of images from DNA gel electrophoresis.
  • Colonoscopy image analysis for 3D surface reconstruction.

More detailed list

Small illustrative image Small illustrative image Small illustrative image Small illustrative image Small illustrative image Small illustrative image Small illustrative image Small illustrative image Small illustrative image Small illustrative image Small illustrative image Small illustrative image Small illustrative image

Funding

We are or have been funded by the Czech ministeries of Education and Health, Czech Grant Agency, Grant Agency of the Czech Academy of Sciences, and european projects; we also have direct collaboration with industry.

Collaborations

  • EPFL, Switzerland
  • INSA, Lyon, France
  • INRIA Sophia-Antipolis, France
  • UPM, Madrid, Spain
  • CIMA, Pamplona, Spain
  • DkfZ, Heidelberg, Germany
  • TUM, Mnichov, Germany
  • STI Medical, U.S.A
  • University of Iowa, U.S.A
  • 1.LF UK, Praha, Czech Republic
  • Masarykova Univerzita, Brno, Czech Republic

Selected journal publications

Current list of publications