Advertisement

Recognition of Neoplastic Changes in Digital Images of Exfoliated Nuclei of Urinary Bladder – A New Approach to Classification Method

  • Annamonika Dulewicz
  • Adam Jóźwik
  • Paweł Jaszczak
  • Bogusław D. Piȩtka
Part of the Advances in Intelligent and Soft Computing book series (AINSC, volume 57)

Summary

The aim of this study was to examine whether it could be possible to recognize neoplastic changes in digital images of exfoliated nuclei of urinary bladder with the help of pattern recognition methods. Nonsurpervised classification based on the k-nearest neighbors rule (k-NN) was applied. Presence of neoplastic urothelial nuclei in organic fluid points to neoplastic changes. A computer-assisted system for identification of neoplastic urothelial nuclei was constructed [2]. The system analyzed Feulgen stained cell nuclei obtained with bladder washing technique and analysis was carried out by means of a digital image processing system designed by the authors. Features describing nuclei population were defined and measured. Then a multistage classifier was constructed to identify positive and negative cases [3]. In this study we used the same features and tried to use k-NN rule to classified analyzed cases. At the beginning the training set was formed of 55 cases representing 19 healthy persons and 36 cancer patients, among them 17 being diagnosed as having cancer of high grade malignancy and 19 as having cancer of low grade malignancy. Standard and parallel k-NN classifiers were analyzed. For both methods feature selection was performed and a total error rate was calculated. Then evaluation of both classifiers was carried out by the leave out method. The evaluation of the examined classifying methods was completed on a set of 76 new cases of testing set. The results for the standard k-NN classifier was: 64% specificity and 77% sensitivity for all the cancer cases. The results for the parallel k-NN classifier was: 57% specificity and 62,5% sensitivity. Then the final approach was carried on join data: training and testing. What made together 131cases. This number of data became numerous enough for analyzing k-NN classiffiers for 8 variables. Finally, the results for the standard k-NN classifier was: 74% specificity and 75% sensitivity for all the cancer cases. The results for the parallel k-NN classifier was: 89% specificity and 74% sensitivity. The results shown that k-NN parallel classifier is sufficiently effective to be used in constructed computer-aided systems dedicated for aid in detection of urinary bladder cancer.

Keywords

Feature Selection Urinary Bladder Method Feature Selection Pattern Recognition Method Neoplastic Change 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Boon, M.E., Drijver, J.S.: Routine cytological staining techniques. Theoretical Background and Practice. Macmillan Education Ltd., London (1986)Google Scholar
  2. 2.
    Dulewicz, A., Piȩtka, D., Jaszczak, P., Nechay, A., Sawicki, W., Pykało, R., Koźmińska, E., Borkowski, A.: Computer identification of neoplastic urothelial nuclei from the bladder. Analytical and Quantitative Cytology and Histology 23(5), 321–329 (2001)Google Scholar
  3. 3.
    Piȩtka, D., Dulewicz, A., Jaszczak, P.: Pathology explorer (PATHEX) a computer- aided system for urinary bladder cancer detection. In: XIII Scientific Conference, Biocybernetics and Biomedical Engineering, Gdańsk, CD-ROM Proceedings, Session XII-2 (2003)Google Scholar
  4. 4.
    Jóźwik, A., Vernazza, G.: Recognition of leucocytes by a parallel k-NN classifiers, Warsaw. Lecture Notes of ICB Seminar, pp. 138–153 (1988)Google Scholar
  5. 5.
    Jóźwik, A., Serpico, S., Roli, F.: A parallel network of modified 1-NN and k-NN classifiers -application to remote-sensing image classification. Pattern Recognition Letters 19, 57–62 (1998)CrossRefzbMATHGoogle Scholar
  6. 6.
    Kurzyński, M.: Rozpoznawanie obrazów, Oficyna Wydawnicza Politechniki Wrocławskiej, Wrocław, str.12–45, 58–102, 143–216 (1997)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2009

Authors and Affiliations

  • Annamonika Dulewicz
    • 1
  • Adam Jóźwik
    • 2
  • Paweł Jaszczak
    • 1
  • Bogusław D. Piȩtka
    • 1
  1. 1.Institute of Biocybernetics and Biomedical Engineering PAS Email: ibib@ibib.waw.plWarsawPoland
  2. 2.Computer Engineering DepartmentTechnical University of ŁódźŁódźPoland

Personalised recommendations