| Summary | The bootstrap is a method for statistical inference from observations that does not necessarily require a parametric model. For instance, this approach allows one to associate a standard deviation to quantities for which there is no theoretical formula giving the standard deviation.
We are currently studing the value of the bootstrap approach to solve a number of problem related to quantification in SPECT and PET. One of our first results was to demonstrate that the bootstrap approach allows one to characterize the statistical properties of SPECT and PET images whatever the statistical properties of the projections and the reconstruction algorithm. The bootstrap thus makes it possible to associate a standard deviation to any measurement performed on a SPECT or PET image |
| Fundings | |
| Publications | - I. Buvat. A non-parametric bootstrap approach for analysing the statistical properties of SPECT and PET images. Phys. Med. Biol., 47:1761-1775, 2002. Article (PDF)
- I. Buvat, C. Riddell. A bootstrap approach for analyzing the statistical properties of SPECT and PET images. IEEE Med. Imaging Conference. San Diego, 2001. Poster (PDF) - I. Buvat, S. Hapdey, C. Riddell, M. Soret. Noise properties of reconstructed SPECT images as a function of the reconstruction scheme. J. Nucl. Med. 42, 138P, 2001. Communication orale (PDF) - I. Buvat, C. Riddell, M. Soret, S. Hapdey. Accuracy of standard errors associated with region of interest measurements in quantitative PET. J. Nucl. Med. 42, 101P, 2001. - I. Buvat, H. Benali, R. Di Paola. Characterizing the noise properties of reconstructed emission images from a single noisy sinogram using a bootstrap approach. J. Nucl. Med, 41, 101P, 2000.Communication orale (PDF) |
| More... | A bootstrap tutorial Slides in French (PDF) |
| Contacts | Irène Buvat : buvat@imed.jussieu.fr |