Classification of Prostate Cancerous Tissues by Support Vector Machine Algorithm with Different Kernels from T2-Weighted Magnetic Resonance Images

Document Type : Original Article (s)

Authors

1 MSc Student, Department of Medical Physics, School of Medicine, Isfahan University of Medical Sciences, Isfahan, Iran

2 Assistant Professor, Department of Medical Physics, School of Medicine, Isfahan University of Medical Sciences, Isfahan, Iran

3 Radiologist, Department of Radiology, Asgariyeh Hospital, Isfahan, Iran

4 Professor, Department of Pathology, School of Medicine, Isfahan University of Medical Sciences, Isfahan, Iran

5 Professor, Department of Medical Physics, School of Medicine, Isfahan University of Medical Sciences, Isfahan, Iran

Abstract

Background: Prostate cancer is one of the most prevalent cancer types in Iran and worldwide. Prostate cancer imaging had been promoted using magnetic resonance imaging (MRI). The aim of present study was to estimate prostate tumors volume by a computerized approach.Methods: By using a Matlab command, the regions of interest were precisely identified. The Haralick features were applied. In addition, using the principal component analysis algorithm, five important features were selected among 17 features. Then, a support vector machine classifier was applied to classify cancerous and normal tissues. To increase the accuracy of the machine vector classifier, the proposed solutions were applied: 1) a new feature was introduced and extracted, 2) all features were normalized, 3) to optimize mutual validation, k-fold changed from 5 to 10. In addition, the support vector machine classifier was implemented by using the Gaussian kernel, radial basis function, and linear kernel. If the tumor was identified in more than one slice, all identified region of interest (ROIs) in different slices were considered in the feature extractions and tumor volume estimation processes.Findings: Among the Haralick features, contrast, correlation, homogeneity, energy, and entropy were the most powerful features in this study that confirmed the findings of previous studies. The sensitivity of the classifier was obtained 0.9180 using Gaussian kernel, while with radial basis function and linear kernels obtained 0.7097 and 0.8571, respectively. In addition, the specificity of Gaussian, radial basis function, and linear kernels were obtained 0.6500, 0.8305, and 0.7069, respectively. The accuracy with Gaussian and linear kernels was obtained 0.7851 which was greater than with the radial basis function. The feature extraction of the regions of interest, feature reduction, and classification steps took less than one minute which indicated the proposed algorithm was fast. It was also repeatable.Conclusion: The proposed computerized estimation of prostate tumors volume can increase the accuracy of the diagnosis. It is quick and simply repeatable.

Keywords


  1. Liu L, Tian Z, Zhang Z, Fei B. Computer-aided detection of prostate cancer with MRI: Technology and applications. Acad Radiol 2016; 23(8): 1024-46.
  2. Puech P, Rouviere O, Renard-Penna R, Villers A, Devos P, Colombel M, et al. Prostate cancer diagnosis: Multiparametric MR-targeted biopsy with cognitive and transrectal US-MR fusion guidance versus systematic biopsy--prospective multicenter study. Radiology 2013; 268(2): 461-9.
  3. Lawrence EM, Tang SY, Barrett T, Koo B, Goldman DA, Warren AY, et al. Prostate cancer: performance characteristics of combined T(2)W and DW-MRI scoring in the setting of template transperineal re-biopsy using MR-TRUS fusion. Eur Radiol 2014; 24(7): 1497-505.
  4. Artan Y, Haider MA, Langer DL, van der Kwast TH, Evans AJ, Yang Y, et al. Prostate cancer localization with multispectral MRI using cost-sensitive support vector machines and conditional random fields. IEEE Trans Image Process 2010; 19(9): 2444-55.
  5. Hambrock T, Vos PC, Hulsbergen-van de Kaa CA, Barentsz JO, Huisman HJ. Prostate cancer: Computer-aided diagnosis with multiparametric 3-T MR imaging--effect on observer performance. Radiology 2013; 266(2): 521-30.
  6. Vos PC, Barentsz JO, Karssemeijer N, Huisman HJ. Automatic computer-aided detection of prostate cancer based on multiparametric magnetic resonance image analysis. Phys Med Biol 2012; 57(6): 1527-42.
  7. Palisaar RJ, Graefen M, Karakiewicz PI, Hammerer PG, Huland E, Haese A, et al. Assessment of clinical and pathologic characteristics predisposing to disease recurrence following radical prostatectomy in men with pathologically organ-confined prostate cancer. Eur Urol 2002; 41(2): 155-61.
  8. Villers A, Puech P, Mouton D, Leroy X, Ballereau C, Lemaitre L. Dynamic contrast enhanced, pelvic phased array magnetic resonance imaging of localized prostate cancer for predicting tumor volume: Correlation with radical prostatectomy findings. J Urol 2006; 176(6 Pt 1): 2432-7.
  9. Lemaitre L, Puech P, Poncelet E, Bouye S, Leroy X, Biserte J, et al. Dynamic contrast-enhanced MRI of anterior prostate cancer: Morphometric assessment and correlation with radical prostatectomy findings. Eur Radiol 2009; 19(2): 470-80.
  10. Haralick RM. Statistical and structural approaches to texture. Proceedings of the IEEE 1979; 67(5): 786-804.
  11. Wibmer A, Hricak H, Gondo T, Matsumoto K, Veeraraghavan H, Fehr D, et al. Haralick texture analysis of prostate MRI: Utility for differentiating non-cancerous prostate from prostate cancer and differentiating prostate cancers with different Gleason scores. Eur Radiol 2015; 25(10): 2840-50.
  12. Gnep K, Fargeas A, Gutierrez-Carvajal RE, Commandeur F, Mathieu R, Ospina JD, et al. Haralick textural features on T2 -weighted MRI are associated with biochemical recurrence following radiotherapy for peripheral zone prostate cancer. J Magn Reson Imaging 2017; 45(1): 103-17.
  13. Smith LI. A Tutorial on Principal Components Analysis (Computer Science Technical Report No. OUCS-2002-12) [Online]. [cited 2002]; Available from: URL: http://hdl.handle.net/10523/7534
  14. Citak-Er F, Vural M, Acar O, Esen T, Onay A, Ozturk-Isik E. Final Gleason score prediction using discriminant analysis and support vector machine based on preoperative multiparametric MR imaging of prostate cancer at 3T. Biomed Res Int 2014; 2014: 690787.
  15. Jayachandran A, Dhanasekaran R. multi class brain tumor classification of MRI images using hybrid structure descriptor and Fuzzy logic based RBF kernel SVM. Iran J Fuzzy Syst 2017; 14(3): 41-54.
  16. Litjens G, Debats O, Barentsz J, Karssemeijer N, Huisman H. Computer-aided detection of prostate cancer in MRI. IEEE Trans Med Imaging 2014; 33(5): 1083-92.
  17. May M, Siegsmund M, Hammermann F, Loy V, Gunia S. Visual estimation of the tumor volume in prostate cancer: a useful means for predicting biochemical-free survival after radical prostatectomy? Prostate Cancer Prostatic Dis 2007; 10(1): 66-71.
  18. Vanitha CDA, Devaraj D, Venkatesulu M. Gene expression data classification using support vector machine and mutual information-based gene selection. Procedia Comput Sci 2015; 47: 13-21.
  19. Chang CY, Hu HY, Tsai YS. Prostate cancer detection in dynamic MRIs. Proceedings of the 2015 IEEE International Conference on Digital Signal Processing, DSP 2015; 2015 Jul 21-24; Singapore.
  20. Li J, Weng Z, Xu H, Zhang Z, Miao H, Chen W, et al. Support Vector Machines (SVM) classification of prostate cancer Gleason score in central gland using multiparametric magnetic resonance images: A cross-validated study. Eur J Radiol 2018; 98: 61-7.
  21. Donati OF, Afaq A, Vargas HA, Mazaheri Y, Zheng J, Moskowitz CS, et al. Prostate MRI: evaluating tumor volume and apparent diffusion coefficient as surrogate biomarkers for predicting tumor Gleason score. Clin Cancer Res 2014; 20(14): 3705-11.
  22. Morgan VA, Parker C, MacDonald A, Thomas K, deSouza NM. Monitoring tumor volume in patients with prostate cancer undergoing active surveillance: Is MRI apparent diffusion coefficient indicative of tumor growth? AJR Am J Roentgenol 2017; 209(3): 620-8.