Advanced

Detectability of fiducial markers in an MRI-only radiotherapy treatment workflow for prostate

Emin, Sevgi LU (2017) FYSK02 20171
Department of Physics
Abstract
In a conventional prostate cancer radiotherapy treatment workflow, computed tomography (CT) and magnetic resonance imaging (MRI) are used for radiotherapy treatment planning and target delineation. MRI is necessary in the process due to better soft tissue contrast compared to CT-images. Using two different imaging modalities in radiotherapy treatment planning requires image registration, which could lead to inaccuracies in calculations. Therefore, an MRI-only based workflow for prostate cancer treatment has been suggested.

Gold fiducial markers (GFMs) are inserted in the prostate prior to treatment. GFMs are used for positioning of the prostate during treatment. GFMs may also be used for image registration between CT- and MR- images in... (More)
In a conventional prostate cancer radiotherapy treatment workflow, computed tomography (CT) and magnetic resonance imaging (MRI) are used for radiotherapy treatment planning and target delineation. MRI is necessary in the process due to better soft tissue contrast compared to CT-images. Using two different imaging modalities in radiotherapy treatment planning requires image registration, which could lead to inaccuracies in calculations. Therefore, an MRI-only based workflow for prostate cancer treatment has been suggested.

Gold fiducial markers (GFMs) are inserted in the prostate prior to treatment. GFMs are used for positioning of the prostate during treatment. GFMs may also be used for image registration between CT- and MR- images in a conventional workflow.

The fiducial markers create a high intensity signal in CT- images, but only a signal void in MRI. The identification method of GFMs in MR-images is currently a manual process requiring manual labor and therefore extra resources. An automatic identification method for GFMs has therefore been developed in a previous study. In this method, a new sequence of MR-images was used, referred to as multi-echo gradient echo images (MEGRE). The MEGRE image sequence was verified for clinical application by a human observatory study.

The human observatory study was conducted using T2-weighted (T2w) MR- and MEGRE images of 44 prostate cancer patients.
The purpose of the study was to compare the performance of human observers and the automatic method in identification of gold fiducial markers to assess the accuracy of the methods.

A matrix laboratory (MATLAB) graphical user interface (GUI) was created for the human observatory study. The GUI was based on an existing GUI for the automatic GFM identification method. The study was conducted at the Department of Medical Physics at Skåne University Hospital in Lund.

Three observers conducted the study and the performance was evaluated for two different conditions. For the first condition, a true marker position was considered to be within the distance of 7.5 mm from the reference coordinates, whilst for the second one the distance was 5 mm. The results of the human observatory study showed 98.33 ± 1.53 %, 95.33 ± 4.16 %, 97.33 ± 2.31 % and 96.00 ± 1.73 %, 89.67 ± 5.69 %, 94.33 ± 3.21 % for the sensitivity, specificity and accuracy at 7.5 mm and 5 mm respectively. The internal differences average between the markers selected by the observers compared to reference markers was 1.07 ± 1.04 mm for T2w images.

The MEGRE image sequence was validated during the study and the results showed a better accuracy in MEGRE images compared to T2w images. The performance of the human observatory study in terms of sensitivity, specificity and accuracy was superior to the automatic method. However, deviations between observers existed, which is expected to be avoided by using the automatic method. (Less)
Please use this url to cite or link to this publication:
author
Emin, Sevgi LU
supervisor
organization
course
FYSK02 20171
year
type
M2 - Bachelor Degree
subject
keywords
prostate cancer, fiducial markers, CT, MRI, MRI-only workflow
language
English
id
8917343
date added to LUP
2017-06-22 11:48:30
date last changed
2017-06-22 11:48:30
@misc{8917343,
  abstract     = {In a conventional prostate cancer radiotherapy treatment workflow, computed tomography (CT) and magnetic resonance imaging (MRI) are used for radiotherapy treatment planning and target delineation. MRI is necessary in the process due to better soft tissue contrast compared to CT-images. Using two different imaging modalities in radiotherapy treatment planning requires image registration, which could lead to inaccuracies in calculations. Therefore, an MRI-only based workflow for prostate cancer treatment has been suggested.

Gold fiducial markers (GFMs) are inserted in the prostate prior to treatment. GFMs are used for positioning of the prostate during treatment. GFMs may also be used for image registration between CT- and MR- images in a conventional workflow.

The fiducial markers create a high intensity signal in CT- images, but only a signal void in MRI. The identification method of GFMs in MR-images is currently a manual process requiring manual labor and therefore extra resources. An automatic identification method for GFMs has therefore been developed in a previous study. In this method, a new sequence of MR-images was used, referred to as multi-echo gradient echo images (MEGRE). The MEGRE image sequence was verified for clinical application by a human observatory study.
 
The human observatory study was conducted using T2-weighted (T2w) MR- and MEGRE images of 44 prostate cancer patients. 
The purpose of the study was to compare the performance of human observers and the automatic method in identification of gold fiducial markers to assess the accuracy of the methods. 

A matrix laboratory (MATLAB) graphical user interface (GUI) was created for the human observatory study. The GUI was based on an existing GUI for the automatic GFM identification method. The study was conducted at the Department of Medical Physics at Skåne University Hospital in Lund. 

Three observers conducted the study and the performance was evaluated for two different conditions. For the first condition, a true marker position was considered to be within the distance of 7.5 mm from the reference coordinates, whilst for the second one the distance was 5 mm. The results of the human observatory study showed 98.33 ± 1.53 %, 95.33 ± 4.16 %, 97.33 ± 2.31 % and 96.00 ± 1.73 %, 89.67 ± 5.69 %, 94.33 ± 3.21 % for the sensitivity, specificity and accuracy at 7.5 mm and 5 mm respectively. The internal differences average between the markers selected by the observers compared to reference markers was 1.07 ± 1.04 mm for T2w images. 

The MEGRE image sequence was validated during the study and the results showed a better accuracy in MEGRE images compared to T2w images. The performance of the human observatory study in terms of sensitivity, specificity and accuracy was superior to the automatic method. However, deviations between observers existed, which is expected to be avoided by using the automatic method.},
  author       = {Emin, Sevgi},
  keyword      = {prostate cancer,fiducial markers,CT,MRI,MRI-only workflow},
  language     = {eng},
  note         = {Student Paper},
  title        = {Detectability of fiducial markers in an MRI-only radiotherapy treatment workflow for prostate},
  year         = {2017},
}