Browse Prior Art Database

Efficient Method for Detection of Lung Diaphragm in MV Fluoroscopic Images

IP.com Disclosure Number: IPCOM000203913D
Original Publication Date: 2011-Feb-09
Included in the Prior Art Database: 2011-Feb-09
Document File: 6 page(s) / 385K

Publishing Venue

Siemens

Related People

Juergen Carstens: CONTACT

Abstract

In External Beam Radiation Therapy (EBRT), monitoring the organ movement during treatment is important for precise treatment delivery. In the case of lung treatment, this is challenging because the organ is continuously moving due to patient breathing. In LINear particle ACcelerators (LINAC), the Mega Volt (MV) treatment beam can be used for imaging the patient during treatment, on the Electronic Portal Imaging Device (EPID). For certain lung tumors, the lung diaphragm movement has direct correspondence with that of the tumor movement and it is adequate to track the diaphragm movement. Also the MV image gives the real time state of the lung. However, the problems associated with diaphragm detection are that the diaphragm is not a rigid object because it undergoes shape changes, and the MV image from the EPID system has pulsing artifacts that affect reliable detection. Pulsing artifacts are due to non-synchronism of image read outs and pulse-timings of the beam. Figure 1 shows a typical image with pulsing artifacts. Furthermore, the detection needs to be robust. The existing methods are computationally intensive.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 38% of the total text.

Page 01 of 6

(This page contains 01 pictures or other non-text object)

(This page contains 00 pictures or other non-text object)

Efficient Method for Detection of Lung Diaphragm in MV Fluoroscopic Images

Idea: Manivannan Sundarapandian, IN-Bangalore

In External Beam Radiation Therapy (EBRT), monitoring the organ movement during treatment is

important for precise treatment delivery. In the case of lung treatment, this is challenging because the

organ is continuously moving due to patient breathing. In LINear particle ACcelerators (LINAC), the

Mega Volt (MV) treatment beam can be used for imaging the patient during treatment, on the

Electronic Portal Imaging Device (EPID). For certain lung tumors, the lung diaphragm movement has

direct correspondence with that of the tumor movement and it is adequate to track the diaphragm

movement. Also the MV image gives the real time state of the lung.

However, the problems associated with diaphragm detection are that the diaphragm is not a rigid

object because it undergoes shape changes, and the MV image from the EPID system has pulsing

artifacts that affect reliable detection. Pulsing artifacts are due to non-synchronism of image read outs

and pulse-timings of the beam. Figure 1 shows a typical image with pulsing artifacts. Furthermore, the

detection needs to be robust. The existing methods are computationally intensive.

There are methods available to address these problems in isolation. For example, non-registration

methods are available for detecting shape changes. Gradient magnitude based techniques are used in

existing proposals. However, this cannot be applied for images with pulsing artifacts as explained.

Filtering techniques are available to suppress pulsing artifacts that remove periodic noise. These

methods are computationally intensive.

The proposed solution is a combination of two preprocessing steps followed by a shape matching

technique. In the full EPID image, the user defines a broader Region Of Interest (ROI) in the

application, e.g. by specification of the diaphragm region of the lung. Since the lung has breathing

movement, it is not to be recommended to get a very narrow ROI from the user input. It is assumed

that the user specifies a square ROI of around one-fourth of the image size to start with. This is

referred as a sub-image. The algorithm operates on the sub-image and consists of the following steps:
• Gradient extraction and threshold
• Search window limitation
• Hough transform
Gradient extraction:

Generally, for computing the Hough transform of the image, the edge information of geometric objects

are used. In most of the shape matching techniques, the gradient magnitude is usually employed to

get the edge information. In the case of MV images with pulsing artifacts the gradient magnitude will

deteriorate the image as illustrated in Figure 6 and 7.

Since the pulsing artifacts appear as periodic vertical strips, the absolute of Y-gradient Sobel operator

is proposed to extract...