1 / 18

The use of PCA-based Methods in the Design of Normal Spect rCBF atlases.

The use of PCA-based Methods in the Design of Normal Spect rCBF atlases. Charlotte Bjuren Supervisors: Prof. Alex Houston Prof. Peter Hancock. Aim of PhD.

teigra
Télécharger la présentation

The use of PCA-based Methods in the Design of Normal Spect rCBF atlases.

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. The use of PCA-based Methods in the Design of Normal Spect rCBF atlases. Charlotte Bjuren Supervisors: Prof. Alex Houston Prof. Peter Hancock

  2. Aim of PhD. • The aim of my PhD is to investigate the potential of a multivariate method for computer aided detection of abnormalities in SPECT imaging • We know that most brains are physiologically different; using a standardised template as found brain imaging software only corrects for anatomical differences. • To address this problem a procedure has been proposed for producing normal atlases, that involves registering and normalising normal images and extracting the mean image plus an appropriate number of normal variants (Eigen images) using Principal Components Analysis Postgraduate Presentation

  3. Medical Image Format MRI Images SPECT Images Postgraduate Presentation

  4. Single Photon Emission Computed Tomography Postgraduate Presentation

  5. Postgraduate Presentation

  6. How Images are Processed. • Registration to template, (alignment of images) are done in Brass, or Multi-Modality. • Brass automatically fits images to an HMPAO templates, created from 35 individuals. • Displays Defect and Difference Images. • Multi-Modality can load up to 3 studies from different modalities. (Pet, CT, MRI & SPECT). • A good visualisation software is Hybrid Viewer. Postgraduate Presentation

  7. EFFECT OF IMAGE REGISTRATION ON A SINGLE TRANSVERSE SLICE IMAGE SLICE TEMPLATE SLICE REGISTERED IMAGE SLICE MEAN IMAGE SLICE Postgraduate Presentation

  8. Can’t tell individual from Mean Image! Postgraduate Presentation

  9. MEAN IMAGE Postgraduate Presentation

  10. EIGENIMAGE 1 Postgraduate Presentation

  11. Extraction of Eigen Images. • Eigen vectors are formed by performing PCA on a set of Normal Images using the voxel as a variable. Since these eigenvectors are in the form of images, they are called eigen images. • They represent the ordered normal variants within the normal dataset and will include mainly physiological variation within the normal dataset and statistical noise. Postgraduate Presentation

  12. Optimisation of the number of Eigen images. • The ration of variance contributed by an eigen image to the total variance in the normal dataset is given by its Eigen value.(eigenimages are ordered by their Eigen values) • It maybe assumed that signal-to-noise ratio in a Eigen image will decrease as the eigenvalue decreases. • An indication of how many Eigen images to include in the atlas is provided by a plot of Eigen values.(altough more sophisticated methods exist) Postgraduate Presentation

  13. CUT-OFF AT P = 5 Postgraduate Presentation

  14. SLICE OF NORMAL BRAIN

  15. MEAN SLICE

  16. SLICE OF EIGENIMAGE 1

  17. First Year Project • The first year will involve developing a statistical method for determining the optimal number of Eigen images. • The method will then be applied to HMPAO brain SPECT-imaging. • We will compare various optimisation methods of extracting Eigen images such as an EigenvalueScree Plot vsJacknifing. Postgraduate Presentation

  18. Reference List • Houston, A. S. (1998). Combining cross-validation and jackknifing to assess the validity of a normal brain atlas. In E. H. M. S. Berry (Ed.), (pp. 53-56). University of Leeds. • Houston, A. S., Kemp, P. M., Griffiths, P. T., & MacLeod, M. A. (1994a). An estimation of noise levels in HMPAO RCBF SPECT images using simulation and phantom data; comparison with results obtained from repeated normal controls. Physics in Medicine and Biology, 39, 873-884. • Houston, A. S., Kemp, P. M., & MacLeod, M. A. (1995). How can we define normality in a medical image? In Yves Bizais (Ed.), Information Processing in Medical Imaging: International Conference, Ile De Berder, France, June 1995 14th (Computational Imaging and Vision) (pp. 351-352). Kluwer Academic Publisher. • Houston, A. S., Fleming, J. S., Ward, T., & Hoffmann, S. M. A. (2009). Optimization of the parameters of a method for computer-aided detection of perfusion deficiencies in brain images. Nuclear Medicine Communications, 30. • Charlotte’s Homepage: charlottebjuren.com Postgraduate Presentation

More Related