Performance Evaluation of Pap Smear Cell Image Classification Using Quantitative and Qualitative Features Based on Multiple Classifiers

Performance Evaluation of Pap Smear Cell Image Classification Using Quantitative and Qualitative Features Based on Multiple Classifiers

• By Riana, Dwiza; Murni, Aniati;
• Seminar Internasional
• 142
• 38-42

Abstract— This paper presents the results of a study on Pap smear cell image classification using qualitative features, as an effort to improve the results of the previous study using quantitative features especially in the performance of the case using all(7)-class category of diseases.  Three good classifiers have been chosen and they include the Naïve Bayes (NB), Multi-Layer Perceptrons (MLPs), and Decision Tree learning algorithm (J48). Twenty quantitative features and seven categories of Pap smear cell image class have been used in this study. Three classes of which are normal cell image class categories that include: Normal Superficial, Normal Intermediate, and Normal Columnar, and the other four classes are categories of abnormal cell image class that include: Mild (Light) Dyplasia, Moderate Dysplasia, Severe Dysplasia and Carcinoma In Situ.  The conversion of quantitative to qualitative features and the image classification process were done with the aid of the Weka software.  The experimental study shows that in all(7)-class classification based on the qualitative features and using the three classifiers with the Weka Correctly Classification Instances (CCI) and Kappa Coefficient classification performance measures, the Naïve Bayes classifier performs the best with the CCI of 73.33% and the Kappa Coefficient of 0.69.  If it is compared to the results of using the quantitative features, the Multiple Layer Perception classifier has performed the best with the CCI of 71.43% and the Kappa Coefficient of 0.67, while the Naïve Bayes produced the CCI of 66.67% and the Kappa Coefficient of 0.61.  The use of qualitative features can improve the performance of the CCI approximately 2% to 7% and of the Kappa coefficient approximately 0.02 to 0.08. 

• Katakunci
• URL http://
• Penerbit ,

REFERENSI

[1] J. Jantzen, J. Norup, G. Dounias, and B. Bjerregaard, Papsmear Benchmark Data For Pattern Classification, Technical University of Denmark, Denmark, 2005.

[2] Martin, Erik. Pap-Smear Classification. Technical University of Denmark – DTU.2003. http://fuzzy.iau.dtu.dk/download/martin2003

[3] J. Norup, Classification of Pap smear Data by Transductive Neuro-Fuzzy Methods, Master Thesis, Technical University of Denmark, Denmark, 2005.

[4] Amalia, Ratih.  Pap Smear Cell Image Classification Using Adaptive Network-based Fuzzy Inference System (ANFIS) and Biomedical Image Retrieval System. Final Project. Faculty of Computer Science, University of Indonesia, (2007)

[5] Giri, Endang P. Pap Smear Image Classification Based On Association Rules for Biomedical Image Retrieval System. Thesis. Faculty of  Computer Science, University of Indonesia, (2008)

[6] Baykut, A. and Ercil. A. Towards Automated Classifier Combination for Pattern Recognition. Multiple Classifier Systems, Springer Verlag, 2003, T. Wideatt, Fabio Roli (eds.), (2003) 94-105 

[7] Saerens, M. and Fouss. F. Yet, Another Method for Combining Classifiers Outputs: A Maximum Entropy Approach. In Proc. of MCS’04, the 5th International Workshop on Multiple Classifier Systems, LNCS 3077, (2004) 82-91

[8] Zhang, B. and Srihari, S.N. Class-Wise Multi-Classifier Combination Based on Dempster-Shafer Theory. In Proc. of ICARV’02, (2002)

[9]  Bi, Y., Bell, D., Wang, H., Guo, G. and Greer, K. Combining Multiple Classifiers Using Dempster-Shafer’s Rule for Text Categorization. In Proc. of MDAI’04, LNCS 313/2004,(2004) pp.127-138

[10] Xu, L., Krzyzak, A. and Suen, C. Methods of Combination Multiple Classifiers and Their Applications to Handwritten Recognition, IEEE Transactions on Systems, Man and Cybernetics, SMC-22(3): 418-435, May/June 1992.

[11] Nadal, C., Legault, R. and Suen, C. Y. Complementary Algorithms for the Recognition of Totally Unconstrained Hand Written Numeral. In Proc. of the 10th International Conference on Pattern Recognition, Volume A (1990) 434-449

[12] Ho, T.K., Hull, J.J. and Srihari, S. N. Decision Combination in Multiple Classifier Systems, IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume 16,Issue 1, (1994) 66 – 75

[13] Dwiza, R. and Murni, A. Classification of Pap Smear Cell Image Based On Quantitative Features Using Multiple Classifier System, International Seminar on ‘IT for Pride and Wealth of Nation’, 25 November 2009 (accepted).

[14] Witten, I.H. and Frank, G. Data Mining: Practical Machine Learning Tools with Java Implementations, Morgan Kaufmann, San Francisco (2000)

[15] Ryan, Shevaun and Hall, Mark. Practical Data Mining, University of Waikato, August 2006.

[16] Daniel Neagu, Gongde Guo, and Shanshan Wang. An Effective Combination Based On Class-Wise Expertise of Diverse Classifiers for Predictive Toxicology Data Mining, Springer-Verlag Berlin Heidelberg, (2006)

[17] Christense, Leslie A and Smith, Robert D, Information System Quality and Value: A Comparison Study of User Versus IS Perception, Journal of Information Technology Management 2 (Number 3), (1999)