4D Image Verification

v ¨ Ubersicht

Auf dem Gebiet der Bestrahlungstherapie gab es in den letzten Jahren weitreichende Entwick- lungen und Fortschritte. Eine besondere Bedeutung auf diesem Gebiet ist der Bestrahlung von Tumoren im Brust- und Bauchbereich zuzuschreiben. In diesen Regionen werden sowohl die Positionen von Organen als auch von Tumorgewebe innerhalb des K¨ orpers signifikant durch die Atmung des zu behandelnden Patienten beeinflusst. Die daraus resultierende Bewegung des Tumorgewebes wird durch eine Erweiterung der zu bestrahlenden Fl¨ ache kompensiert. Diese Erweiterung beinhaltet alle m¨ oglichen Positionen des Tumors und erstreckt sich somit auch auf gesundes Gewebe.

Mehrere klinische Studien belegen eine h¨ ohere Erfolgsquote bei erh¨ ohter Strahlendosis. Um m¨ oglichst wenig gesundes Gewebe einer solch hohen Strahlung auszusetzen verwenden ¨ Artze

das Konzept der ’gated radiotherapy’. Derzeit g¨ angige Ans¨ atze basieren auf der ¨ Uberpr¨ ufung eines Ersatzsignales. Dies kann sowohl ein implantierter Marker, als auch ein externes Signal sein, welches versucht die Atmung des Patienten zu erfassen.

Innerhalb dieser Arbeit werden Grundlagen und Methoden von ’gated radiotherapy’ erl¨ autert. Zus¨ atzlich wird eine ¨ Ubersicht ¨ uber aktuelle Patente und Produkte auf diesem Gebiet gegeben, sowie Vor- und Nachteile derzeit g¨ angiger Ans¨ atze diskutiert.

Basierend auf dieser Diskussion wird eine neu entwickelten Methode vorgestellt. Diese vereint die Vorteile beider bereits bekannten Ans¨ atze, vermeidet dabei jedoch deren Nachteile. Der hierf¨ ur entwickelte Algorithmus bedient sich bildbasierter Verfahren und Methoden der medizinischen Bildverarbeitung. Er berechnet eine Zuordnung zwischen einem 4D - CT Plan- nungsvolumen und einer kurz vor der Behandlung aufgenommenen R¨ ontgenbildsequenz des- selben Patienten. Mit Hilfe dieser Zuordnung ist es dem Arzt m¨ oglich auf Basis eines exter- nen Atemsignals Grenzen f¨ ur die Bestrahlung zu definieren und den Patienten nur in speziellen Phasen der Atmung zu bestrahlen.

Der entwickelte Algorithmus ist in einen bereits existierenden medizinischen Prototypen in- tegriert, entwickelt von Siemens Corporate Research (SCR) in Princeton, NJ, USA. Mit Hilfe dieses Prototyps wird die Anwendbarkeit der Methode demonstriert. Zudem ist ein weiterer Prototyp zur Aufnahme von R¨ ontgenbildsequenzen, synchronisiert mit einem Atemsignal, en- twickelt worden. Beide Applikationen werden innerhalb dieser Arbeit vorgestellt.

vi

Abstract

Far reaching developments and technical advances took place within the field of radiotherapy in the last years. Radiotherapy within the chest and abdomen area is especially important in the field of radiotherapy. Within this regions, organ and tumor positions are significantly affected by patient respiration. The tumor motion, caused due to respiration is compensated by extending the treated area. This extension covers all possible positions of the tumor and therefore also includes healthy tissue.

Several clinical studies provide evidence of a survival advantage for higher dose levels. To spare a maximum of healthy tissue physicians use ’gated radiotherapy’. Common recent ap- proaches for gated radiotherapy are based on the observation of a surrogate. This either can be an implanted fiducial marker or an external signal, which is trying to capture the patients’ respiration.

Within this thesis principles and methods of ’gated radiotherapy’ are described. Additionally an overview of recent patents and products related to radiotherapy are presented and advantages and disadvantages of both common approaches are discussed. This discussion leads to a new developed method, which is introduced. The method joins advantages of both known methods but disregards their disadvantages. The developed algorithm is using image guided methods and methods of medical image processing. A mapping between a 4D-CT planning volume and a most recent acquired fluoroscopic sequence of the same patient is calculated before treatment. Using this mapping and an external breathing signal the physician can define gating intervals and treat the patient in certain breathing phases.

The developed algorithm is included in an existing prototype developed by Siemens Corpo- rate Research (SCR) in Princeton, NJ, USA. Using this prototype, the application of the method is shown. Furthermore another prototype to acquire respiration synchronized fluoroscopic se- quences is developed. Both applications are introduced within this thesis.

viii

Abbreviations

2D Two Dimensional

3D Three Dimensional

4D Four Dimensional

4D-CT Four Dimensional Computed Tomography

AAPM American Association of Physicists in Medicine

ACS American Cancer Society

AP Anterior - Posterior

CCD Charge Coupled Device

CT Computed Tomography

CTV Clinical Target Volume

DDR Digital Reconstructed Radiography

DICOM Digital Imaging and Communications in Medicine

DP Dynamic Programming

EM Expectation Maximization

EPID Electronic Portal Imaging Device

FDA Food and Drug Administration

GTV Gross Tumor Volume

IGRG Image Guided Respiration Gating

IMRT Intensity Modulated Radiation Therapy

IMRS Intensity Modulated Radiosurgery

IRB Institutional Review Board

IRIS Integrated Radiology Information System

ISD Inverse of the Standard Deviation

ITV Internal Target Volume

LINAC Linear Accelerator

MGH Massachusetts General Hospital

MIP Maximum Intensity Projection

MRI Magnetic Resonance Imaging

PTV Planning Target Volume

RANSAC RANdom SAmple Consensus

RPM Real-time Position Management

RTRT Real Time Tumor Tracking

SI Superior - Inferior

ix

Abbreviations

SVD Singular Value Decomposition

XML eXtensible Markup Language

Contents

1  Introduction  1

2  Related Work and Patents  5

3  Managing Respiratory Motion in Radiation Therapy  9

3.1  Introduction  9

3.2  Treatment Planning  12

3.3  Motion-encompassing methods  13

3.3.1  Slow CT scanning  14

3.3.2  Inhalation and exhalation breath-hold CT  14

3.3.3  Four-dimensional CT respiration-correlated CT  14

3.4  Respiratory Gating Methods and Procedures  15

3.4.1  Internal Gating  15

3.4.2  External Gating  17

3.5  Clinical Procedure  18

4  Image Sequence Synchronization  23

4.1  Introduction  23

4.2  Proposed Method  24

4.3  Similarity Matrix  26

4.4  Preprocessing  28

4.4.1  Gain Removal  28

4.4.2  Wild Card  30

4.4.3  Transition matrix  34

4.5  Model-based Dynamic Programming  36

xi

  xii CONTENTS  

5  Clinical Prototype  43

5.1  Introduction  43

5.2  Clinical Systems  44

5.2.1  SOMATOM Sensation Open  44

5.2.2  ONCOR Linear Accelerator  45

5.2.3  Respiratory Gating System  46

5.3  Clinical Applications  48

5.3.1  AcquireIt  48

5.3.2  RTReg4D  50

6  Results  55

6.1  Introduction  55

6.2  Level I: Synthetic Data  57

6.3  Level II: Phantom Data  58

7  Discussion and Future Work  69

7.1  Introduction  69

7.2  Wild Card Detection Improvement  69

7.3  Model Improvement  70

7.4  Cone Beam Acquisition Integration  70

7.5  On-the-fly Expansion  71

7.6  Registration Improvement  72

7.7  Parallelization  72

8  Summary  73

List of Figures 77

List of Tables 81

  Bibliography  83

Chapter 1

Introduction

Lung cancer is still the most fatal type of cancer, even though the cases of death have been declined in the last years. Regarding to the American Cancer Society (ACS), lung cancer was the reason for almost 29% of all cancer deaths in the United States in 2005. There were 172,570 new estimated cases diagnosted with an estimated 163,510 deaths. Although one can recognize a continuous downward movement of deaths caused by lung cancer in the past 16 years, it is important to improve early detection and treatment continously (Figure 1.1).

Besides radiation therapy lung cancer is also treated with surgery, chemotherapy, and targeted biological therapies, depending on the type and stage of the cancer. Regarding to the ACS, the 1- year relative survival rate for lung cancer was 42% in 2000, however the 5-year relative survival rate for all stages combined was only 15% [Ame05].

Receiving radiation treatment therapy especially in the thorax and abdomen area causes a major health risk for patients. As the tumor is moving due to intrafraction organ motion which is mainly caused by patient respiration, physicians have to treat healthy tissue as well. Facing this movement, physicians are inevitable stuck in their decision about a proper treatment for the patient. On the one hand to be able to deliver x-rays throughout the whole treatment physicians have to increase the treated region by the whole range, where the tumor potentially could be. On the other hand there is clinical evidence of survival advantage for higher dose levels [Oku95], [Per86] [Mac05]. Concerning these two conflictive issues, physicians are biased. Increasing the dose level would accord a better chance of survival, but regarding the extended treatment region, a lot of healthy tissue would be exposed to this high dose, too. Certainly, this is very bad for the patient.

An optimal designed system should provide the radiation oncologist both opportunities. The oncologist still should be able to use a higher dose for treatment, but at the same time minimal

1

2 CHAPTER 1. INTRODUCTION

Figure 1.1: Age-Adjusted Cancer Death Rates, US, 1930 - 2002; Source: US Mortality Public Use Data Tapes 1960 - 2002, US mortality Volumes 1930-1959, National Center for Health Statistics, Centers for Disease Control and Prevention, 2004

healthy tissue should be harmed. Therefore a system should both consider patient respiration and enable a significant decrease of the treated area.

About 15 years ago, physicians came up with the idea of using gated radiotherapy to face this problem [Oha89]. The general idea of gated radiotherapy is to reduce the incidence and severity of normal tissue complications and to increase local control through dose escalation. In order to obtain these issues, a range of values within the treatment beam is turned on has to be specified. Therefore the localization of the tumor has to be known first. This either can be achieved by tracking an implanted fiducial marker or by assuming a correlation between an external surrogate placed on the patient chest for example. Hence, gated radiotherapy is divided into two groups (Figure 1.2):

• internal gating

• external gating

Both methods are based on the observation of a surrogate, where the correlation between an internal surrogate is considered to be more accurate than a correlation between an external surrogate and the real tumor position. Considering these two methods, there are various problems and each method has its own advantages and disadvantages, respectively (Table 1).

3

Figure 1.2: [1]: Internal Gating, a fiducial marker is implanted in the tumor (green) and used as a surrogate; [2]: External Gating, an external surrogate is used, assuming a correlation between the tumor and the surrogate

Internal gating requires additional dose for the patient because image acquisition is necessary to locate the implanted marker properly. This additional dose can be more than what is clinically acceptable for patients with many treatment fractions or a long treatment time of a single fraction. Furthermore internal gating is difficult for thoracic tumors or even not possible because fiducial markers cannot be inserted and there is a risk of pneumothorax.

In opposite external gating is non-invasive but it is quite inaccurate because there often is a bad correlation between the external surrogate and the real tumor position. Consolidating, current state of the art techniques are not satisfying enough in a clinical environment. With this thesis a solution using image guided methods is provided. This solution joins the advantages of internal and external gating and abandons the disadvantages of both techniques (Table 1). The proposed algorithm is computing an image-based mapping immediately before treatment which is based upon a reference breathing cycle and the current breathing of the pa- tient. During the treatment the breathing pattern is observed by an external surrogate which can be compared to the computed mapping. Therefore no surgery is required and the patient is only exposed to a short additional dose which is just used to acquire a few breathing cycles. Further- more the correlation is computed prior the treatment and is not based upon external surrogates, but on patient images where breathing phases and tumor position are known.

To obtain this goal, a general overview of literature related to this topic as well as related

4 CHAPTER 1. INTRODUCTION

Correlation Dose Additional Surgery

4D Image Verification + + +

Table 1.1: Comparison of internal / external gating and 4D Image Verification

patents are presented first. Afterwards basic principles of gated radiotherapy and there fields of application are described. As the therapy is divided into a planning and a treatment stage, the requirements for a planning setup are shown. Furthermore motion-encompassing methods to ac- count tumor motion are discussed and introduced. Based on acquired data using these methods it is possible to apply internal and external gating techniques. Examples for both of them are also presented within this theses. Finally an example for a clinical treatment procedure and guide- lines for gated radiotherapy are introduced. Accordingly the proposed image guided method is discussed in detail and all necessary technical background is introduced. As the developed al- gorithm has to be tested in clinical environment, its implementation into an existing prototype is also described within this theses. Additionally a second prototype to acquire fluoroscopic sequences correlated with a breathing signal is introduced. Medical devices used for data acqui- sition are also introduced, as well as a use case for the application of the prototype. To appraise the quality and ability of the proposed solution various results of performed tests are discussed and proposals for further developments are presented in the end.

Chapter 2

Related Work and Patents

Many studies related to respiratory motion and gated radiotherapy for lung cancer have been published in literature. Since gated radiotherapy is a recent technology, there are still a lot fea- sibility and evaluation studies being performed. This chapter provides an overview of published work within the last years and points at the latest products in the 4-D treatment market which are introduced in detail in the following chapter. Recently, the American Association of Physicists in Medicine (AAPM) has released a report about the management of respiratory motion in radiation oncology [AAP06]. As this report is a recommendation on how to manage respiratory motion, it is a valuable source for background informations about the topic dealt within this theses and therefore taken strongly into consideration within the next chapter.

Applying gated radiotherapy to lung tumors requires some understanding about tumor motion in general. Various evaluations about this issue have been published lately, e.g. [Pla04], [Err03], [Six03], [For03], [Man03], [Sep02], [Ste01], [Shi01], [Gir01], [Ekb98]. Summarizing these studies, the motion magnitude can be clinically significant and depending on tumor sites and individual patients. The range is within the limits of a few centimeters.

Once it is obvious how tumor motion is behaving it is essential to figure out if and how tumor motion is correlated with a respiratory signal. [Mag04], [Ahn04], [Hoi04], [Tsu04], [Koc04], [Ved03] provide an overview of this correlation (Table 2).

Furthermore, it is proven that there is a clinical evidence of survival advantage for higher dose levels [Oku95], [Per86] [Mac05]. To deliver such a highly conformal radiation dose distribution to a complex static target volume physicians can use intensity-modulated radiation therapy in combination with gated radiotherapy (gated IMRT) [Kub00].

Once it is obvious that there is a correlation between breathing and tumor movement, the theoretical background for gated radiotherapy is given [Oha89], [Ort95] (see chapter 3.1). Com-

5

6 CHAPTER 2. RELATED WORK AND PATENTS

Organ/source Respiratory N patients Correlation Phase shift Source

fluoroscopy abdominal

displacement

correlated

CT

Table 2.1: Correlation of tumor/organ motion with the respiratory signal; 3D: three - dimen-

sional, AP: anterior - posterior, CT: computed tomography, MRI: magnetic resonance imaging,

pts: patients, s: second(s), SI: superior - inferior [AAP06]

mon terms for tumor volumes are discussed in two reports of the International Commission on

Radiation Units and Measurements [ICR93], [ICR99] (see chapter 3.1).

Keall performed a theoretical analysis of margins to examine potential reduction of CTV-

PTV margins, which is the main goal of gated radiotherapy [Kea02]. A second study regarding

the reduction of CTV-PTV margins confirms a possible reduction of the margin, using gated

radiotherapy [Bar01] (see chapter 3.1).

These margins are created and defined in a process called treatment planning. Sean provides

principles and guidelines for lung cancer treatment planning [Sen04a], [Sen04b] (see chapter

3.2). The reference sequence used within the treatment planning phase is acquired by using

7

motion encompassing methods (see chapter 3.3). Three of these methods are introduced within this thesis. These methods are slow CT scanning [Lag01] [Kos01], inhale and exhale breath-hold CT [Bal98], [Aru98] and 4D-CT/respiration-correlated CT [Son05] [Lu05] [Pan04] [Rie05]. As mentioned before gated radiotherapy can be divided into two sections, internal gating and external gating. As an example for an internal gating system, the real-time tumor-tracking radio- therapy (RTRT) system developed jointly by Mitsubishi Electronics and the Hokkaido University is introduced [Kun99] [Shi01] [Shi00c] [Shi00a] [Shi03] [Shi99] [Kit02] [Shi00b] (see chapter 3.4.1). Currently another internal gating system is developed by Calypso Medical Technologies which avoids some disadvantages of the RTRT system [Kra02]. External gating systems are provided by Varian Medical Systems [For02] [Mos02], BrainLab [Sch01] and Siemens Medical Solutions (see chapter 3.4.2).

To understand the clinical process of gated radiotherapy an example developed at the Mas- sachusetts General Hospital (MGH) and guidelines from the American Association of Physicists in Medicine (AAPM) are introduced [Jia05] [AAP06] (see chapter 3.5).

The proposed algorithm is embedded into an existing project developed by Siemens Corpo- rate Research [Kha06]. This project comprehends required algorithms for patient positioning and similarity measurements like mutual information [Wel96]. To obtain a solution for the given problem the core algorithm uses Dynamic Programming [Bel57], [Dre77] which is modified by a model-based approach. The underlying model is based upon the principles of markov chains [Mar71]. However, to achieve valuable input data for the core algorithm, some preprocessing steps have to be applied. Within the preprocessing step an existing significant gain field has to be removed, which is done by taking advantage of the singular value decomposition (SVD) [Tre97]. General explanation of markov chains and dynamic programming are based upon [Nie83].

Chapter 3

Managing Respiratory Motion in

Radiation Therapy

3.1 Introduction

Since Wilhelm Conrad Roentgen discovered X-rays in 1895, overwhelming possibilities in can- cer treatment were potentiated. Within the last century and especially recent years this powerful technology was continuously improved [Ort95]. Nowadays radiation oncologists are able to re- lieve pain or even cure cancer by using Roentgens X-rays. However, even after more than 100 years of research there are still problems in delivering X-rays in a way the therapy has the highest probability of cure with the least morbidity at the same time.

One of these problems is respiratory motion of tumor targets. Due to breathing, tumor sites located in the thorax and abdomen are especially affected by respiratory motion. Especially one kind of cancer is extremely difficult to treat, which is lung cancer. The position of tumors located in the patient’s lung is especially affected by respiration during treatment sessions. As respiration is regulated in the medulla by stimulating the diaphragm, this is an ’involuntary’ action the patient cannot prevent, even though breath hold techniques could control breathing in a restricted way.

To facilitate the gas (O 2 - CO 2 ) exchange between blood and air, the diaphragm is contin- uously being contracted and relaxed. This mechanism has a quite significant impact on organs and tumors. General speaking referring to respiratory motion literature, one cannot assume a general respiration pattern for particular patients prior to observation and treatment. Breathing patterns are individual and dissimilar for each patient. Concerning a specific patient, breathing patterns vary not only from one treatment session to another but also even during a single session

9

10 CHAPTER 3. MANAGING RESPIRATORY MOTION IN RADIATION THERAPY

as well [Sep02] [Nei06]. Motion between two sessions is called inter-fraction motion whereas the motion during a treatment session is called intra-fraction motion.

Farther in terms of tumor volumes, there are four terms to address certain tumor and treatment regions in literature [ICR93], [ICR99]:

• Gross Tumor Volume (GTV): is referred to as the tumor itself

• Clinical Target Volume (CTV): GTV extended by the areas of sub-clinical disease

• Planning Target Volume (PTV): internal margin considering the organ movement and pos- sible setup errors are added to the GTV

• Internal Target Volume (ITV): internal target volume incorporating the motion of GTV

As the gross tumor volume is not covering all possible afflicted tissue, the physicians have to extend the volume to cover metastasis and other sub-clinical disease. In order to treat moving tumor targets, all locations of a possible residence of the tumor have be covered as well. However as already stated before, creating and irradiating a PTV leads to a suboptimal solution because a lot of healthy tissue is constantly exposed to high doses. In order to avoid this problem and to improve radiation treatment, physicians can apply respiratory gated radiation therapy. In the late 1980s and early 1990s respiratory gated radiation therapy was developed in Japan first [Oha89]. The technique relies upon a correlation between tumor motion and respiration, using a respiration signal as a surrogate. Irradiation is only performed in a specific state of the breathing cycle, as the tumor is roughly at the same location of the body each time the patient is in a specific breathing cycle state. While the tumor is moving within this specified part of the breathing cycle, it is moving within the so called ’gate’ or ’gating window’ (Figure 3.1). According to this term the name of gated radiotherapy is derived.

There are two different types of gating:

• Displacement Gating:

In order to apply displacement gating, the relative position between two extremes of breath- ing motion, namely, inhalation and exhalation is measured. The beam is activated when- ever the respiration signal is within a range of relative positions.

• Phase Gating Breathing is divided into breathing phases, like begin of inhale, maximum inhale or end of exhale. Often these phases are also referred to percentage levels, like 10% inhale or 10% exhale. Whenever the patient’s respiration is within certain phases, the beam is turned on.

3.1. INTRODUCTION 11

Figure 3.1: Application of a gating window; the tumor (blue) is moving [a], the corresponding breathing curve is shown in [b], and the gating window (beam on) is indicated by the red border

In an idealized gated treatment, tumor position should be directly detected and the delivery of radiation is only allowed when the tumor is at the right position. However the direct detection of the tumor mass in real-time during the treatment often is difficult or even impossible. Therefore a system needs to know the patients’ respiration to verify the predefined gate as well as to observe surrogates, which can either be an external respiration signal or an internal fiducial marker.

Apparently, the duration of a treatment session is extended if gated radiation therapy is ap- plied because the beam is not able to deliver X-rays continuously. But this leads to the primary objective of gated radiation therapy. Treating the tumor just at specific positions opens a poten- tial for a significant decrease of the CTV-PTV margin and therefore tumors can be treated with higher dose levels. These levels are measured in gray units (Gy), which is the absorption of one joule of radiation energy by one kilogram. At the moment, studies on the clinical or dosimetric gain of margin reduction are rare and preliminary. Theoretical analysis of margin reduction by studying phantoms using gated radiotherapy promise a reduction of 2-11 mm in the CTV to PTV margin [Kea02]. Another study showed, that using deep inspiration breath hold techniques on

12 CHAPTER 3. MANAGING RESPIRATORY MOTION IN RADIATION THERAPY

average decreases the percent of lung volume receiving > 20Gy (V20 1 ) from 12.8% by 11% without and by 8.8% with CTV-PTV margin reduction [Bar01]. According to this, gated radio- therapy enables the possiblity of exposing less healthy tissue to high dose levels and therefore the patient has a chance to achieve a better long-term survival rate.

As mentioned before, treatment sessions are obviously consuming more time, because tu- mors are irradiated just within certain breathing phases. The ratio of beam-on time to the total treatment time is referred to as duty cycle and is a measurement of the treatment efficiency. As some tumor motion still occurs within the gate, this motion is also called residual motion. How- ever, there is always a trade-off between duty cycle and residual motion, because increasing the duty cycle results in a larger residual motion of the tumor. Usually both phases are taken into consideration during the treatment planning. According to physicians there is usually a larger duty cycle achieved during exhale than during inhale. Therefore most patients are treated at full exhale if there is a larger duty cycle required. For a more accurate gating and a stable residual motion patients are gated at full inhale, because this breathing phase is more reproducible than full exhale.

In the following, an overview of principles and methods to account for respiratory motion in radiotherapy is given. This section is abutted to the AAPM Task Group 76 report ’The Manage- ment of Respiratory Motion in Radiation Oncology’ [AAP06].

3.2 Treatment Planning

The main objective of defining a CTV-PTV margin is founded in possible geometric and setup errors. Therefore, a planning session is scheduled before the patient is treated. Within this session an acquisition of the target region using one of the below described methods is done. This acquisition is used by physicians to define the CTV-PTV margin. Principles and guidelines for treatment planning regarding lung cancer can be found in two publications by Senan et al. [Sen04a], [Sen04b]. The following components take influence on CTV-PTV margins definitions for lung cancer:

• Inter- and intra-observer variations in GTV delineation:

Due to different experience and education of physicians, target regions are defined different quite often [Gir02]. Therefore a defined GTV region is dependent on the physician and not unique.

1 the volume of lung treated to a dose of = 20Gy