Abstract—In this study, a three stage dimension reduction strategy is proposed for early detection of prostate cancer by using mass spectrometry data. In the initial stage, a filtering method is used. While in the second stage, two different methods namely, the wavelet analysis and statistical moments are used for comparison. The last stage includes a feature transformation method which is called kernel partial least squares algorithm. After dimension reduction stages, prostate mass spectrometry data are classified with k-nearest neighbor, support vector machines and linear discriminant analysis. The classification process is handled in two phases. In the first phase, the prostate mass spectrometry data are classified as the normal and cancerous samples with an accuracy of 95.8%. While in the second phase, the cancerous samples are classified as benign and malign samples with an accuracy of 87.2%. For each cases it is shown that, the combination of the wavelet analysis and kernel partial least squares methods is sufficient for prostate cancer identification.

Index Terms—Classification, kernel partial least squares, mass spectrometry, prostate cancer, wavelet analysis.

The authors are with the Department of Electronics and Communication Engineering, Istanbul Technical University, 34439 Turkey (e-mail: taskinv@ itu.edu.tr, bdogan@ itu.edu.tr, olmezt@ itu.edu.tr).

 

Cite:Vedat Taşkın, Berat Doğan, and Tamer Ölmez, "Prostate Cancer Classification from Mass Spectrometry Data by Using Wavelet Analysis and Kernel Partial Least Squares Algorithm," International Journal of Bioscience, Biochemistry and Bioinformatics vol. 3, no. 2, pp. 98-102, 2013.