An object-oriented library for systematic training and comparison of classifiers for computer-assisted tumor diagnosis from MRSI measurements

  • Frederik O. Kaster
  • Bernd Merkel
  • Oliver Nix
  • Fred A. Hamprecht
Special Issue Paper

Abstract

We present an object-oriented library for the systematic training, testing and benchmarking of classification algorithms for computer-assisted diagnosis tasks, with a focus on tumor probability estimation from magnetic resonance spectroscopy imaging (MRSI) measurements. In connection with a graphical user interface for data annotation, it allows clinical end users to flexibly adapt these classifiers towards changed classification tasks, to benchmark various classifiers and preprocessing steps and to perform quality control of the results. This poses an advantage over previous classification software solutions, which required expert knowledge in pattern recognition techniques in order to adapt them to changes in the data acquisition protocols. This software will constitute a major part of the MRSI analysis functionality of RONDO, an integrated software platform for cancer diagnosis and therapy planning which is under current development.

Keywords

Magnetic resonance spectroscopy imaging Computer-assisted diagnostics Statistical classification Automated quality control 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Arya S, Mount D, Netanyahu N et al. (1998) An optimal algorithm for approximate nearest neighbor searching. J ACM 45:891–923 MATHCrossRefMathSciNetGoogle Scholar
  2. 2.
    Bandos A, Rockette H, Gur D (2007) Exact bootstrap variances of the area under ROC curve. Commun Stat, Theory Methods 36:2443–2461 MATHCrossRefMathSciNetGoogle Scholar
  3. 3.
    Bengio Y, Grandvalet Y (2004) No unbiased estimator of the variance of K-fold cross-validation. J Mach Learn Res 5:1089–1105 MathSciNetGoogle Scholar
  4. 4.
    Breiman L (1996) Out-of-bag estimation. Tech. rep., UC Berkeley Google Scholar
  5. 5.
    Chang C, Lin C (2001) LIBSVM: a library for support vector machines. Software available at http://www.csie.ntu.tw/cjlin/libsvm
  6. 6.
    Cho S, Kim M, Kim H et al. (2001) Chronic hepatitis: in vivo proton MR spectroscopic evaluation of the liver and correlation with histopathologic findings. Radiology 221(3):740–746 CrossRefGoogle Scholar
  7. 7.
    Dager S, Oskin N, Richards T, Posse P (2008) Research applications of magnetic resonance spectroscopy (MRS) to investigate psychiatric disorders. Top Magn Reson Imaging 19(2):81–96 CrossRefGoogle Scholar
  8. 8.
    Demšar J (2006) Statistical comparisons of classifiers over multiple data sets. J Mach Learn Res 7:1–30 MathSciNetGoogle Scholar
  9. 9.
    Dietterich T (1998) Approximate statistical tests for comparing supervised classification learning algorithms. Neural Comput 10:1895–1923 CrossRefGoogle Scholar
  10. 10.
    de Edelenyi FS, Rubin C, Estève F et al. (2000) A new approach for analyzing proton magnetic resonance spectroscopic images of brain tumors: nosologic images. Nat Med 6:1287–1289 CrossRefGoogle Scholar
  11. 11.
    Fawcett T (2006) An introduction to ROC analysis. Pattern Recognit Lett 27(8):861–874 CrossRefMathSciNetGoogle Scholar
  12. 12.
    Frigo M, Johnson S (2005) The design and implementation of FFTW3. Proc IEEE 93(2):216–231 CrossRefGoogle Scholar
  13. 13.
    García-Gomez J, Luts J, Julià-Sapé M et al. (2009) Multiproject-multicenter evaluation of automatic brain tumor classification by magnetic resonance spectroscopy. Magn Reson Mater Phys 22:5–18 CrossRefGoogle Scholar
  14. 14.
    Gillies R, Morse D (2005) In vivo magnetic resonance spectroscopy in cancer. Annu Rev Biomed Eng 7:287–326 CrossRefGoogle Scholar
  15. 15.
    Golub G, Heath M, Wahba G (1979) Generalized cross-validation as a method for choosing a good ridge parameter. Technometrics 21(2):215–223 MATHCrossRefMathSciNetGoogle Scholar
  16. 16.
    González-Vélez H, Mier M, Julià-Sapé M et al. (2009) HealthAgents: distributed multi-agent brain tumor diagnosis and prognosis. Appl Intell 30:191–202 CrossRefGoogle Scholar
  17. 17.
    Görlitz L, Menze B, Weber M et al. (2007) Semi-supervised tumor detection in magnetic resonance spectroscopic images using discriminative random fields. In: Proceedings DAGM 2007. Lecture notes in computer science, vol 4713/2007, pp 224–233 Google Scholar
  18. 18.
    de Graaf R (2008) In vivo NMR spectroscopy: principles and techniques. Wiley, New York Google Scholar
  19. 19.
    Grandvalet Y, Bengio Y (2006) Hypothesis testing for cross-validation. Tech Rep TR 1285, Département d’Informatique et Recherche Opérationelle, University of Montréal Google Scholar
  20. 20.
    Hagberg G (1998) From magnetic resonance spectroscopy to classification of tumors: a review of pattern recognition methods. NMR Biomed 11(4–5):148–156 CrossRefGoogle Scholar
  21. 21.
    Hastie T, Tibshirani R, Friedman J (2009) The elements of statistical learning. Springer, New York MATHCrossRefGoogle Scholar
  22. 22.
    Kaster F, Kelm B, Zechmann C et al. (2009) Classification of spectroscopic images in the DIROlab environment. In: World congress on medical physics and biomedical engineering, IFMBE Proc, vol 25/V, pp 252–255 Google Scholar
  23. 23.
    Kelm B, Menze B, Neff T et al. (2006) CLARET: a tool for fully automated evaluation of MRSI with pattern recognition methods. In: Handels H, Ehrhardt J, Horsch A et al. (eds) Bildverarbeitung für die Medizin 2006 – Algorithmen, Systeme, Anwendungen, pp 51–55 CrossRefGoogle Scholar
  24. 24.
    Kelm B, Menze B, Zechmann C et al. (2007) Automated estimation of tumor probability in prostate magnetic resonance spectroscopic imaging: pattern recognition vs quantification. Magn Reson Med 57:150–159 CrossRefGoogle Scholar
  25. 25.
    Köthe U (2000) Generische Programmierung für die Bildverarbeitung. Ph.D. thesis, University of Hamburg, software available at http://hci.iwr.uni-heidelberg.de/vigra/
  26. 26.
    Kreis R (2004) Issues of spectral quality in clinical 1H magnetic resonance spectroscopy and a gallery of artifacts. NMR Biomed 17(6):361–381 CrossRefGoogle Scholar
  27. 27.
    Lin H, Lin C, Weng R (2007) A note on Platt’s probabilistic outputs for support vector machines. Mach Learn 68:267–276 CrossRefGoogle Scholar
  28. 28.
    Martínez-Bisbal M, Celda B (2009) Proton magnetic resonance spectroscopy imaging in the study of human brain cancer. Q J Nucl Med Mol Imaging 53(6):618–630 Google Scholar
  29. 29.
    Maudsley A, Darkazanli A, Alger J et al. (2006) Comprehensive processing, display and analysis for in vivo MR spectroscopic imaging. NMR Biomed 19(4):492–503 CrossRefGoogle Scholar
  30. 30.
    Menze B, Lichy M, Bachert P et al. (2006) Optimal classification of long echo time in vivo magnetic resonance spectra in the detection of recurrent brain tumors. NMR Biomed 19(5):599–609 CrossRefGoogle Scholar
  31. 31.
    Menze B, Kelm B, Weber M et al. (2008) Mimicking the human expert: pattern recognition for an automated assessment of data quality in MR spectroscopic images. Magn Reson Med 59:1457–1466 CrossRefGoogle Scholar
  32. 32.
    Neuter BD, Luts J, Vanhamme L et al. (2007) Java-based framework for processing and displaying short-echo-time magnetic resonance spectroscopy signals. Comput Methods Programs Biomed 85:129–137 CrossRefGoogle Scholar
  33. 33.
    Ortega-Martorell S, Olier I, Julià-Sapé M et al. (2010) SpectraClassifier 1.0: a user friendly, automated MRS-based classifier-development system. BMC Bioinform 11:106 CrossRefGoogle Scholar
  34. 34.
    Poullet J, Sima D, Van Huffel S (2008) MRS signal quantitation: a review of time- and frequency-domain methods. J Magn Reson 195(2):134–144 CrossRefGoogle Scholar
  35. 35.
    Provencher S (2001) Automatic quantitation of localized in vivo 1H spectra with LCModel. NMR Biomed 14(4):260–264 CrossRefGoogle Scholar
  36. 36.
    Rifkin R, Klautau A (2004) In defense of one-vs-all classification. J Mach Learn Res 5:101–141 MathSciNetGoogle Scholar
  37. 37.
    Sajja B, Wolinsky J, Narayana P (2009) Proton magnetic resonance spectroscopy in multiple sclerosis. Neuroimaging Clin N Am 19(1):45–58 CrossRefGoogle Scholar
  38. 38.
    Smith S, Levante T, Meier B et al. (1994) Computer simulations in magnetic resonance: an object-oriented programming approach. J Magn Reson A106(1):75–105 CrossRefGoogle Scholar
  39. 39.
    Stefan D, Cesare FD, Andrasescu A et al. (2009) Quantitation of magnetic resonance spectroscopy signals: the jMRUI software package. Meas Sci Technol 20:104 035 CrossRefGoogle Scholar
  40. 40.
    Stroustrup B (2001) Exception safety: concepts and techniques. In: Dony C, Knudsen J, Romanovsky A et al. (eds) Advances in exception handling techniques. Springer, New York, pp 60–76 CrossRefGoogle Scholar
  41. 41.
    Tate A, Underwood J, Acosta D et al. (2006) Development of a decision support system for diagnosis and grading of brain tumours using in vivo magnetic resonance single voxel spectra. NMR Biomed 19(4):411–434 CrossRefGoogle Scholar
  42. 42.
    Xu D, Vigneron D (2010) Magnetic resonance spectroscopy imaging of the newborn brain—a technical review. Semin Perinatol 34(1):20–27 CrossRefGoogle Scholar
  43. 43.
    Zechmann C, Menze B, Kelm B, Zamecnik P, Ikinger U, Waldherr R, Delorme S, Hamprecht F, Bachert P (2010) How much spatial context do we need? Automated versus manual pattern recognition of 3D MRSI data of prostate cancer patients. NMR Biomed (submitted) Google Scholar
  44. 44.
    Zhu G, Smith D, Hua Y (1997) Post-acquisition solvent suppression by singular-value decomposition. J Magn Reson 124:286–289 CrossRefGoogle Scholar

Copyright information

© Springer-Verlag 2010

Authors and Affiliations

  • Frederik O. Kaster
    • 1
    • 2
  • Bernd Merkel
    • 3
  • Oliver Nix
    • 2
  • Fred A. Hamprecht
    • 1
  1. 1.Heidelberg Collaboratory for Image ProcessingUniversity of HeidelbergHeidelbergGermany
  2. 2.German Cancer Research CenterHeidelbergGermany
  3. 3.Fraunhofer MeVis Institute for Medical Image ComputingBremenGermany

Personalised recommendations