Please use this identifier to cite or link to this item:
Title: Bio-medical X-ray imaging with Synchrotron Radiation: study and implementation of algorithms related to phase sensitive techniques
Authors: Mohammadi, Sara
Supervisore/Tutore: Tromba, Giuliana
Parmigiani, Fulvio
Issue Date: 11-Apr-2013
Publisher: Università degli studi di Trieste
Phase-contrast X-ray imaging is an invaluable tool in medical diagnostics and biological sciences. It provides images where both absorption and refraction contribute. For quantitative analysis of these images, the phase needs to be retrieved numerically. There are many phase-retrieval methods available.
My thesis aims to optimize the application of phase retrieval methods for different phase-contrast imaging situations. A quantitative comparison between phase contrast and phase retrieved images is also performed on some selected examples.

We analyzed the two most-applied phase retrieval algorithms and outlined derivations, approximations and assumptions of each one. We implemented these algorithms on experimental data collected at the SYRMEP beam-line of ELETTRA, Italy. We used the algorithms based on Born approximation and transport-of-intensity equation (TIE) for different kinds of test objects and biological samples (high absorbing, homogeneous, low absorbing, etc.) at various experimental conditions (sample-to-detector distances, energies, detector resolutions). Their capability and restriction are evaluated in the visualization of the sample morphology up to its tiniest details.
It was shown that Born method is highly depended on phase and intensity distributions in the object plane, being quite accurate in the case of small phase variations and low absorption materials. On the other side, the TIE approximation did not impose any direct limitations on phase and intensity distributions in the object-plane or image-plane but its application is restricted to near-filed region.
The application of phase retrieval techniques on biological samples have been assessed the strong ability of these algorithms in improving the visualization of structures and details with a complex geometry. They also produced images with a better separation of the different phases of the sample for an easier and more efficient segmentation. On the other side, as expected, the application of phase retrieval algorithms on high absorbing sample didn’t give further useful information respect to the phase contrast imaging.
The Signal to Noise Ratio (SNR) values obtained for phase retrieved images, either for the area-signal and edge-signal, were significantly improved with respect to conventional phase contrast images. For both algorithms, the noise has been significantly reduced. In general, the application of TIE method produced higher SNR in comparison to Born as it is not strongly limited by absorption of the sample. We were able to reduce the scattering contributions by increasing the sample-to-detector distance and, therefore, higher SNR values were achieved. Using phase retrieval algorithms, the tiny details were less defined than in the original phase contrast image because of blurring, but, on the final analysis, this was not significant since data were rendered and segmented more efficiently.
For the future development of this work, we will consider the possibility to implement a mixed TIE+Born approach in order to overcome the limitation of Born approximation about the requirement for sample absorption, and restriction posed by TIE to the near-field region sample-to-detector working distances.
Ciclo di dottorato: XXV Ciclo
metadata.dc.subject.classification: SCUOLA DI DOTTORATO DI RICERCA IN FISICA
Keywords: phase contrast imaging
phase retrieval
synchrotron radiation
Language: en
Type: Doctoral Thesis
Settore scientifico-disciplinare: FIS/03 FISICA DELLA MATERIA
NBN: urn:nbn:it:units-9990
Appears in Collections:Scienze fisiche

Files in This Item:
File Description SizeFormat
PhD_Thesis_Sara_Mohammadi.pdf17.31 MBAdobe PDFThumbnail
Show full item record

CORE Recommender

Page view(s)

checked on May 31, 2019


checked on May 31, 2019

Google ScholarTM


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