On the self-regularization property of the EM algorithm for Poisson inverse problems

  • Axel Munk
  • Mihaela Pricop


One of the most interesting properties of the EM algorithm for image reconstruction from Poisson data is that, if initialized with a uniform image, the first iterations improve the quality of the reconstruction up to a point and it deteriorates later dramatically. This ’self- regularization’ behavior is explained in this article for a very simple noise model.We further study the influence of the scaling of the kernel of the operator involved on the total error of the EM algorithm. This is done in a semi- continuous setting and we compute lower bounds for the L1 risk. Numerical simulations and an example from fluorescence microscopy illustrate these results.


Positron Emission Tomography Inverse Problem Expectation Maximization Expectation Maximization Algorithm Linear Inverse Problem 
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.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.



M. Pricop and A. Munk acknowledge support of DFG SFB 755 and FOR 916.We are grateful to P. Marnitz, A. Esner, S. Mek and A. Schoenle for the 4π images used in the introduction.


  1. Ahn, S., Fessler, J.A. (2003). Globally convergent image reconstruction for emission tomography using relaxed ordered subsets algorithms, IEEE Transactions on Medical Imaging 22: 613- 626.CrossRefGoogle Scholar
  2. Anderson, J.M.M.,Mair, B. A.&Rao,M. (1996). Positron emission tomography: a review, Control and Cybernetics 25: 1089–1111.MATHMathSciNetGoogle Scholar
  3. Antoniadis,A.&Bigot, J. (2006). Poisson inverse problems, TheAnnals of Statistics 34: 2132–2158.MATHCrossRefMathSciNetGoogle Scholar
  4. Bertero, M. & Boccacci, P. (1998). Introduction to inverse problems in imaging, Bristol.Google Scholar
  5. Bertero, M., Boccacci, P., Desiderà G. & Vicidomini G. (2009). Image deblurring with Poisson data: from cells to galaxies, Inverse Problems, to appear.Google Scholar
  6. Bissantz, N., Hohage, T., Munk, A. & Ruymgaart, F. (2007) Convergence rates of general regularization methods for statistical inverse problems and applications, SIAM J. Numer. Anal. 45: 2610–2636.MATHCrossRefMathSciNetGoogle Scholar
  7. Bissantz, N., Mair, B.A. & Munk, A. (2005). A multi-scale stopping criterion for MLEM reconstruction in PET, IEEE Nuclear Science Symposium Conference Record 6: 3376–3379.Google Scholar
  8. Bissantz, N.,Mair, B.A. & Munk, A. (2008). A statistical stopping rule for MLEM reconstructions in PET, it IEEE Nucl. Sci. Symp. Conf. Rec. 8: 4198-4200.Google Scholar
  9. Bühlmann, P. & Yu, B. (2003). Boosting with the L 2 loss: regression and classification, J. Amer. Statist. Assoc. 98: 324–339.MATHCrossRefMathSciNetGoogle Scholar
  10. Cavalier, L. & Koo, J.-Y. (2002). Poisson intensity estimation for tomographic data using a wavelet shrinkage approach, Transactions on Information Theory 48: 2794–2802.MATHCrossRefMathSciNetGoogle Scholar
  11. Chang, I. S. & Hsiung, C. A. (1994). Asymptotic consistency of the maximum likelihood estimate in positron emission tomography and applications, The Annals of Statistics 22: 1871–1883.MATHCrossRefMathSciNetGoogle Scholar
  12. Coakley, K.J. (1991). A Cross-Validation Procedure for stopping the EM algorithm and deconvolution of neutron depth profiling spectra, IEEE Transactions on Nuclear Science 38: 9–15.CrossRefGoogle Scholar
  13. Coakley, K.J. (1996). Bootstrap method for nonlinear filtering of EM-ML reconstructions of PET images, International Journal of Imaging Systems and Technology 7: 54-61.CrossRefGoogle Scholar
  14. Coakley, K.J. & Llacer, J. (1991). The use of Cross- Validation as a stopping rule in emission computed tomography image reconstruction, SPIE Proc. Image Phys: Med. Imaging V 1443: 226–233.Google Scholar
  15. Cover, T. M. (1984). An algorithm for maximizing expected log investment return Transactions on Information Theory 30: 369–373.MATHCrossRefMathSciNetGoogle Scholar
  16. Crauel, Hans (2002). Random probability measures on Polish spaces, Taylor & Francis, London.MATHGoogle Scholar
  17. Csiszár, I. & Tusnády, G. (1984). Information geometry and alternating minimization procedures, Statistics & Decisions 1: 205–237.Google Scholar
  18. Dempster, A. P., Laird, N. M. & Rubin, D. B. (1977). Maximum likelihood from incomplete data via the EM algorithm, Journal of the Royal Statistical Society. Series B 39: 1–38.MATHMathSciNetGoogle Scholar
  19. Eggermont, P. P. B. (1993). Maximum entropy regularization for Fredholm integral equations of the first kind, SIAM Journal on Mathematical Analysis 24: 1557–1576.MATHCrossRefMathSciNetGoogle Scholar
  20. Eggermont, P. P. B. (1999). Nonlinear smoothing and the EM algorithm for positive integral equations of the first kind, Applied Mathematics and Optimization 39: 75–91.MATHCrossRefMathSciNetGoogle Scholar
  21. Eggermont, P. P. B. & LaRiccia, V. N. (1996). Maximum penalized likelihood estimation and smoothed EMalgorithms for positive integral equations of the first kind, Numerical Functional Analysis and Optimization 17: 737–754.MATHCrossRefMathSciNetGoogle Scholar
  22. Eggermont, P. P. B. & LaRiccia, V. N. (2001). Maximum penalized likelihood estimation. Vol. I, Springer-Verlag, New York.Google Scholar
  23. Engl, H. W. , Hanke, M. & Neubauer, A. (1996). Regularization of inverse problems, 375. Kluwer Academic Publishers Group, Dordrecht.MATHGoogle Scholar
  24. Fessler, J.A. (1994). Penalized weighted least-squares image-reconstruction for positron emission tomography, IEEE Transactions on Medical Imaging 13: 290–300.CrossRefGoogle Scholar
  25. Hebert, T.J. (1990). Statistical stopping criteria for iterative maximum-likelihood reconstruction of emission images, Physics in Medicine and Biology 35: 1221–1232.CrossRefGoogle Scholar
  26. Hebert, T., Leahy, R. & Singh, M. (1988). Fast MLE for SPECT using an intermediate polar representation and a stopping criterion, IEEE Transactions on Nuclear Science 35: 615–619.CrossRefGoogle Scholar
  27. Hudson, M. & Larkin, R. (1994). Accelerated Image Reconstruction using Ordered Subsets of Projection Data, IEEE Trans. Med. Imag 13: 601–609.CrossRefGoogle Scholar
  28. Iusem, Alfredo N. (1992). A short convergence proof of the EM algorithm for a specific Poisson model, Rebrape 6: 57–67.MATHMathSciNetGoogle Scholar
  29. Johnson, V.E. (1994) A note on stopping rules in EM-ML reconstructions of ECT images, IEEE Transactions on Medical Imaging 13, 569-571.CrossRefGoogle Scholar
  30. Johnstone, I. M. & Silverman, B.W. (1990). Speed of estimation in positron emission tomography and related inverse problems, The Annals of Statistics 18: 251–280.MATHCrossRefMathSciNetGoogle Scholar
  31. Koo, J.-Y. & Chung, H.-Y. (1998). Log-density estimation in linear inverse problems, The Annals of Statistics 26: 335–362.MATHCrossRefMathSciNetGoogle Scholar
  32. Kondor, A. (1983).Method of convergent weights—An iterative procedure for solving Fredholm’s integral equations of the first kind, Nuclear Instruments andMethods in Physics Research 216: 177-181.CrossRefGoogle Scholar
  33. Kontaxakis, G. & Tzanakos, G. (1993). Further study of a stopping rule for the EM algorithm, Bioengineering Conference, 1993., Proceedings of the 1993 IEEE Nineteenth Annual Northeast Volume , 18-19: 52 - 53.Google Scholar
  34. Korostelëv, A. P & Tsybakov, A. B. (1993). Minimax theory of image reconstruction, Springer- Verlag, New York.MATHGoogle Scholar
  35. Lang, M. & Müller, T. & Engelhard, J. & Hell, S. (2007). 4Pi microscopy of type A with 1-photon excitation in biological fluorescence imaging, Opt. Express 15: 2459.CrossRefGoogle Scholar
  36. Latham, G. A.& Anderssen, R. S. (1994). On the stabilization inherent in the EMS algorithm, Inverse Problems. 10, 161–183.MATHCrossRefMathSciNetGoogle Scholar
  37. Llacer, J.,Veklerov, E.&Coakley, K.J. (1993). Statistical analysis ofmaximum-likelihood estimator images of human brain FDGPET studies, IEEE Transactions onMedical Imaging 12: 215-231.CrossRefGoogle Scholar
  38. Mair, B. A, Rao, M. & Anderson, J. M. M. (1996). Positron emission tomography, Borel measures and weak convergence, Inverse Problems 12: 965–976.MATHCrossRefMathSciNetGoogle Scholar
  39. Meinshausen, N., Rice, J. & Schücker, T. (2006). Testing for monotonicity in the Hubble diagram, eprint arXiv:astro-ph/0612556 12.Google Scholar
  40. Mülthei,H.N.&Schorr, B. (1989).On properties of the iterativemaximumlikelihood reconstruction method, Mathematical Methods in the Applied Sciences 11: 331–342.MATHCrossRefMathSciNetGoogle Scholar
  41. Natterer, F. (2001). The mathematics of computerized tomography, Philadelphia.Google Scholar
  42. Resmerita, E., Engl, H. W. & Iusem, A. N. (2007). The expectation-maximization algorithm for ill-posed integral equations: a convergence analysis, Inverse Problems 23: 2575–2588.MATHCrossRefMathSciNetGoogle Scholar
  43. Resmerita, E., Engl, H. W. & Iusem, A. N. (2008). Corrigendum: The expectation-maximization algorithm for ill-posed integral equations: a convergence analysis, Inverse Problems 24: 1p.CrossRefMathSciNetGoogle Scholar
  44. Shepp, L.A. & Vardi, Y. (1982). Maximum Likelihood Reconstruction for Emission Tomography, IEEE Transactions on In Medical Imaging 1: 113–122.CrossRefGoogle Scholar
  45. Silverman, B.W., Jones,M. C., Nychka, D.W.&Wilson, J. D. (1990). A smoothed EMapproach to indirect estimation problems, with particular reference to stereology and emission tomography, Journal of the Royal Statistical Society. Series B. 52: 271–324.MATHMathSciNetGoogle Scholar
  46. Szkutnik, Z. (2000).Unfolding intensity function of a Poisson process inmodels with approximately specified folding operator, Metrika 52: 1–26.MATHCrossRefMathSciNetGoogle Scholar
  47. Vardi, Y. & Lee, D. (1993). From image deblurring to optimal investments: maximum likelihood solutions for positive linear inverse problems. With discussion. J. Roy. Statist. Soc. Ser. B 55: 569–612.MATHMathSciNetGoogle Scholar
  48. Vardi,Y., Shepp, L. A.&Kaufman, L. (1985). Astatisticalmodel for positron emission tomography, Journal of the American Statistical Association 80: 8–37.MATHCrossRefMathSciNetGoogle Scholar
  49. Veklerov,E.&Llacer, J. (1987). Stopping rule for theMLEalgorithmbased on statistical hypothesistesting, IEEE Transactions on Medical Imaging 6: 313-319 .CrossRefGoogle Scholar
  50. Vicidomini, G., Hell, S. & Schönle, A. Automatic deconvolution of 4Pi-microscopy data with arbitrary phase, Opt. Lett., Submitted.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2010

Authors and Affiliations

  1. 1.Institut für Mathematische Stochastik, Georg August Universität GöttingenGöttingenGermany

Personalised recommendations