Please use this identifier to cite or link to this item:
Title: Image registration of radiographic images using an elastic approach
Authors: Abche, Antoine 
Tohme, Elie Toni
Chaer, Toufic Naji
Karam, Elie 
Hamam, Yskandar
Bouchoucha, Michel
Rocaries, Francois
Affiliations: Department of Electrical Engineering 
Department of Electrical Engineering 
Keywords: Minimisation
Affine transformation
Diagnostic radiography
Medical image processing
Subjects: Image registration
Issue Date: 2007
Publisher: IEEE
Part of: IEEE Nuclear Science Symposium Conference Record, 2006
Start page: 208
End page: 2085
Conference: IEEE Nuclear Science Symposium Conference Record (29 Oct.-1 Nov. 2006 : San Diego, CA, USA) 
In this work, an elastic registration technique is implemented to align different radiographic images to track the movement of gas and determine the colonic response to food in patients with functional abdominal bloating. The registration transformation is made of two components: i) an affine transformation parameterized in terms of 5 parameters (1 rotation, 2 magnification and 2 translation parameters) and ii) an elastic component expressed in terms of radial functions representing the smoothing term. Then, a chi2 function defined in terms of the sum of the squared residual distances of corresponding internal point landmarks after they are transformed to the same coordinate system, is formed. Thus, the registration can be achieved by "tuning" the parameters until the residuals become simultaneously small (minimum). Since the parameters of the transformation are related in a non-linear fashion, they are estimated iteratively using a non-linear optimization technique, namely, the Simplex method developed by Nelder and Mead. Finally, the results are evaluated visually and quantitatively to study the precision of the registration.
Ezproxy URL: Link to full text
Type: Conference Paper
Appears in Collections:Department of Electrical Engineering

Show full item record

Record view(s)

checked on Jun 4, 2023

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.