Graduation project in collaboration with Dr. Emile A. Hendriks
Information and Communication Theory Group, Technical University Delft.
February 2002 - October 2002
M.W. Jobse, J. Davelaar, B.C. Stoel
In the treatment of patients with cancer, radiotherapy plays an important role. X-rays with an energy of approximately 6 - 10 MeV ('mega volt photons') destroy cells at the location of the tumor, which brings the unrestrained cell growth of the tumor to a standstill. Such a treatment period takes about 30 days, with one radiation treatment per day.
|The radiation, generated by a linear accelerator, must be directed to the tumor very accurately, since it should be prevented that the surrounding healthy tissue would be damaged as well. To determine these target areas, the tumor and surrounding critical organs are made visible using Computed Tomography (CT) images.||
Digital Reconstructed Radiograph (DRR) image for the planning of the radiation therapy.
From these image data, the required optimal radiotherapy beam geometry is calculated using dedicated planning software. This beam geometry is obtained by adjusting the position of tungsten leaves that collimate the X-ray beam to form the specified shape.
|In order to perform quality control, an image is acquired of the projection of the radiotherapy beam during the therapy session. In this image, the radiation field is depicted along with some anatomical structures in low contrast. This mega volt (or portal) image can be compared with a simulated projection which is reconstructed from the CT images ("Digital Reconstructed Radiograph", DRR) by the planning software. By comparing these two images, it is possible to verify the positioning of the patient during radiation therapy and adjust the positioning if necessary.|
The ultimate goal is to analyze these images on-line, in order to adjust the positioning of the patient in real-time. The first step in this development is to develop algorithms that (semi-)automatically matches the two images based on the anatomical structures and calculate the displacement relative to the planned radiation field.
There are mainly two approaches to solve the problem of matching two images in this application: 1) by extracting simple features from the two images (such as contours) and register these features (i.e. chamfer matching); or 2) by using the original gray value information from the two images, and match the images, after a normalization step (i.e. mutual information). Both approaches require pre-processing of the images, as in the first approach the contours should be extracted and in the second the two images should normalized. As the first approach allows user-interaction, this approach is implemented first in this project.
|1. Detection of the borders of the radiation field in the portal image.|
|2. Detection of the leaves in the DRR image, if necessary.|
|3. Matching the radiation fields of the portal image and the DRR image and initialize the displacement measurement.|
|4. Contour detection in both images|
|5. Registration of the two sets of contours
|6. Calculation of the displacement|
Jobse MW, Davelaar J, Hendriks E, Kattevilder R, Reiber JHC, Stoel, BC: A new algorithm for the registration of portal images to planning images in the verification of radiotherapy, as validated in prostate treatments. Med. Phys. 30 (9): 2274-2281, 2003.
Davelaar J, Jobse MW, Hendriks EA, Creutzberg CL, Reiber JHC, Stoel BC, A new algorithm to register portal to planning images in radiotherapy and a validation using prostate treatments European Society for Therapeutic Radiation Oncology (ESTRO), Copenhagen 2003.
For further information, please contact:
Berend C. Stoel, PhD.
Division of Image Processing
Department of Radiology, 1-C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
Tel. +31 (0)71 526 1911
Fax. +31 (0)71 526 6801