Image reconstruction in electrical impedance mammography
Breast cancer continues to be a leading cause of death in Western countries. One of the
most important approaches to reduce this mortality is to detect the cancer as early as
possible. Although current diagnostic imaging modalities are able to give useful
information for diagnosis, development of new imaging technique is highly desirable in
order to detect breast cancer in an earlier stage. This is the motive of the present study of
Electrical Impedance Mammography (ElM), which applies Electrical Impedance
Tomography (EIT) to image human breasts.
The overall aim of the whole project is to develop impedance imaging techniques and
system for early detection of breast cancer. The aim of the studies reported in this thesis is
twofold, to investigate methods of improving EIT image quality and the feasibility of EIT
in breast cancer detection.
Focusing on these aims the following work is reported:
1) The study of three different image reconstruction algorithms is described. In this
work the image reconstruction results are compared and the most appropriate
algorithm was chosen for the subsequent study of breast imaging.
2) An investigation of two important factors in EIT image reconstruction, the number
of electrodes (NOE) and the number of conductivity basis functions (NOCBF)
whose effects on EIT images have yet been studied in detail so far, are described. In
this work the image reconstruction is analysed with different combinations of NOE
and NOCBF using Singular Value Decomposition (SVD) and spectrum expansion
theory. Finally suggestions are given on which configuration could offer better
image quality in breast imaging3) A comprehensive investigation is reported regarding compatibility of different types
of prior information and its effect on an iterative image reconstruction algorithm,
based on which a novel method is proposed to improve EIT image quality. This
method selects compatible prior information by observing the convergence
behaviour of an image reconstruction algorithm. The principle, implementation, and
results are detailed. Results indicate the effectiveness of this method.
4) A two-dimensional breast imaging simulation system is introduced. In this work
several breast models with different physiological and pathological conditions are
made, based on clinical in vitro measurements and Cole-Cole model. Images with
different current frequencies are reconstructed and analysed. Results indicate the
potential of detecting and identifying breast abnormality by EIT.
5) A preliminary study on 3D EIT and 3D electrode placement has been conducted.
The mathematic principle and implementation of 3D EIT are described, followed by
a study on the sensitivity of boundary measurements to the conductivity changes in
a cylindrical object with 2 different types of electrode placement. Suggestions are
given on optimal electrode placement in EIT breast imaging.
Finally suggestions are gIven for future work. These include a) investigating
appropriate electrode placement for different applications and corresponding current &
voltage patterns; b) incorporating more prior information into ElM image reconstruction; c)
designing a more precise 3D breast forward model; and d) investigating appropriate
regularization for image reconstruction