contact    site map  
    polski   english

Image processing and analysis in medicine

Ongoing or completed projects

Quantitative analysis of human bronchial trees with the use of multidetector CT scanner

 

Quantitative analysis of human bronchial trees with the use of multidetector CT scanner

The Institute carries out research in quantitative analysis of human bronchial trees with the use of multidetector CT scanner in the frame of Polish National Science Centre grant nb.  N519 652240 in 2011-2013. The main aim of the project is to develop and test algorithms for processing and analysis of three-dimensional tomographic images of the chest, leading to a quantitative description of the local parameters of bronchial airway i.e. wall thickness measurements and cross-sectional area of ​​the lumen.
 
Bronchial diseases are a major public health problem not only in Poland but also in the World. The most common diseases of bronchi are asthma and chronic obstructive pulmonary disease COPD. Asthma is a chronic inflammatory disorder of the airways in which participate many cells and substances released by them. Asthma attacks is usually accompanied by varying in intensity diffuse obstruction (narrowing) of bronchitis, often subsides spontaneously or with treatment. The frequency of occurrence of asthma increases in all countries, especially among children. Asthma is a significant burden, not only of health care costs, but also lost productivity and reduced participation in family life. It is estimated that asthma affects about 300 million people worldwide. The incidence of asthma in different geographic regions, ranging from 1 to 18%. Approximately, In The world 250 000 people die a year for  asthma (GINA).

Figure 1 presents the steps of processing and tomographic image analysis to determine the local wall thickness and lumen area. A 3D tomographic chest is an input of the algorithm. The first processing step is to remove all the unnecessary parts of the image and leaving the bronchial tree, in other words, it consists in segmentation of the bronchial tree. The next step involves the determination of the skeleton of the tree that is thin curve in the middle of each branch of the bronchi. This skeleton provides in a further step, the construction of cross-sections perpendicular to the skeleton and thus perpendicular to the branches of the bronchi. Determination of perpendicular planes is a necessary condition for making accurate local measurements of wall thickness and bronchial lumen area. The last step is to make mentioned measurements on the cut fragment of  bronchial wall by the plane perpendicular to the skeleton .

Fig 1. Stages of CT image processing and analysis

Completed tasks:
  1. Implementation and testing of a segmentation algorithm of bronchial trees based on region growing in 3D space.
  2. Development, implementation and testing of the algorithm for closing of geometric holes and topological holes.
  3. Development, implementation and testing of the bronchial tree segmentation algorithm using the hole closing procedure. Studies have shown that the algorithm gives better results than the algorithm based on region growing (Postolski, Janaszewski et al. 2009).
  4. Implementation and testing of the following 3D skeletonisation algorithms: a fully parallel thinning algorithm (Ma, Sonka 1996), 12-iterative thinning algorithm (Palagyi, Cuba, 1999), an algorithm based on the theory of cubical complexes  (Chaussard, Couprie 2009), an thinning algorithm based on critical configurations of voxels (Bertrand, Couprie 2009). The best results were achieved for the algorithm based on cubical complexes (Postolski, Janaszewski et al. 2010).
  5. Implementation and testing of algorithms for estimation of  tangent to any point in 3D discrete curve: a) the naive algorithm, b) averaging algorithm, c) curve segmentation algorithm, d) algorithm based on splines, e) λ-MST algorithm in 3D space (original algorithm).
  6. Comparison of algorithms implemented in Section 5 shows that the best results gives λ-MST (Postolski, Janaszewski et al. 2011)
  7. Implementation of the modified version of the half-max algorithm, which is called half-max+. The new algorithm can perform measurements in areas where the half-max model could not be applied.
  8. Development of an algorithm that generates the model segmented bronchi from 3D CT images of a chest. It has been shown that this model is more useful to test the bronchi skeletonisationalgorithms than the earlier model presented in (Kitaoka, Takaki et al., 1999).
Literature
  1. Postolski, M., Janaszewski, M., Fabijańska, A., Babout, L., Couprie, M., Jędrzejczyk, M., Stefańczyk, L.: Reliable Airway Tree Segmentation Based on Hole Closing in Bronchial Walls, Sixth International Conference on Computer Recognition Systems, Jelenia Góra, Poland. Computer Recognition Systems 3. M. Kurzyński,M. Wozniak, Springer p. 389-396, 2009.
  2. Ma, C.M., Sonka, M.: a fully parallel 3D thinning algorithm and its applications. Computer Vision and Image Understanding 64 p. 420–433, 1996.
  3. Palagyi, K., Kuba, A.: a parallel 3D 12-subiteration thinning algorithm. Graphical Models and Image Processing 61, p.  199-221, 1999
  4. Chaussard J, Couprie M.: Surface thinning in 3d cubical complexes, in Proceedings of the 13th International Workshop on Combinatorial Image Analysis, pp. 135-148, 2009.
  5. Bertrand, G, Couprie M.: On parallel thinning algorithms: minimal non-simple sets, P-simple points and critical kernels,”Journal of Mathematical Imaging and Vision, vol. 35, no. 1, pp. 23-35, 2009.
  6. Postolski M., Janaszewski M., Jopek Ł, Babout. L..: 3D skeletonization of pulmonary airway tree structures. Zeszyty Naukowe AGH, Automatyka 14, p. 337-351, 2010.
  7. Postolski M., Janaszewski M., Jopek Ł., Babout L.: Wyznaczanie kierunku stycznej do dowolnego punktu trójwymiarowej krzywej wolumetrycznej w ilościowej analizie ludzkich drzew oskrzelowych. Automatyka 15, p. 219-234, 2011.
  8. H. Kitaoka, R. Takaki, B. Suki, A three-dimensional model of the human airway tree. J Appl physiol 87, 2207, 1999.
Team:
dr inż. M. Janaszewski, IACS, Lodz University of Technology
mgr inż M. Postoslki, IACS, Lodz University of Technology; University Paris East, LIGM, ESIEE, Paris, France
dr hab L. Babout, IACS, Lodz University of Technology
mgr inż L. Jopek, IACS, Lodz University of Technology
Prof L Stefańczyk, Department of Radiology and Diagnostic Imaging, Medical University of Lodz
Lek med. M Jędrzejczyk, Department of Radiology and Diagnostic Imaging, Medical University of Lodz
 
 

The Institute of Applied Computer Science has established colaboration with two medical research institutes: The Research Institute of Polish Mother's Memorial Hospital and The Department of Nephrology and Transplantation Medicine of the Medical University in Wroclaw. The research project carried out at the Department of Nephrology and Transplantation Medicine of the Medical University in Wroclaw is focused on the optimization of the method based on the ELISPOT (enzyme linked immunospot) approach for determination of alloreactivity of renal transplant recipient in clinical practice. The goal of the project is to obtain a non-invasive diagnostic tool for prediction of long-term renal allograft function and early detection of markers of chronic graft rejection process. In the ELISPOT method the image analysis needs evaluation of specific morphological parameters of the objects appearing as round spots. The quantitative analysis of images is possible with the use of commercial systems, but the time and cost of such examination are unacceptable for scientific research. Our department has developed SPOTVIEW - an efficient approach for unattended image segmentation that is based on local threshold supported by a region based method. The results were successfully compared to those obtained with the commercial software. Additionally, as the new method is more sensitive, it enables detection of small, low contrasted spots skipped by a commercial procedure.

The second project, developed for Polish Mother's Memorial Hospital, concerned automatic examination of microscope images of the breast cancer tissue for cancer cell detection and classification. One of the detection techniques is immunohistochemical staining of the tissue. Nuclei with receptors react with a dye and change colors into red-brown, while other remain pale blue. The fraction of colored nuclei in the specific area gives an important hint on choosing proper pharmacological therapy.
To accomplish the goal, which was construction of the system (the name of the project: PATO), we developed the algorithm for color image processing. The algorithm has been constructed with use of standard methods. Some of them needed modification to fulfill the task. These were: new approach to median-like color image filtering, multi-pass binarization algorithm, pixel-classification algorithm and watershed-based region segmentation algorithm.
Both PATO and SPOTVIEW are comparably cheap vision systems. The software doesn't need installation and connection to the apparatus; may be run on every PC, which is very important for the medical research centers.
The computer vision system Pato may be used at any Clinical Pathomorphology Centre for supporting both routinal and scientific experiments on a cancer tissue with immunohistochemical staining.
The system SpotView may be used by transplantology clinics as a valuable tool for monitoring the graft outcome risk by ELISPOT method.
The new variants of the image processing algorithms proposed by the research team may be used for the wide range of computer vision systems: from medical diagnostics, ecology to industry.

The members of the team:
Wojciech Bieniecki PhD
Szymon Grabowski PhD

return to the top


Editors:
Wojciech Bieniecki
Szymon Grabowski

Last modification:
2013-02-19 22:09:15,