Radiation induced lung damage - Chapter 8 Precise and real-time measurement of 3D tumor motion in lung breathing and heartbeat, measured during radiotherapy

UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)

UvA-DARE (Digital Academic Repository)

Radiation induced lung damage

Seppenwoolde, Y.

Publication date

2002

Link to publication

Citation for published version (APA):

Seppenwoolde, Y. (2002). Radiation induced lung damage.

General rights

It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulations

If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.

Chapterr 8

Precisee and real-time measurement of 3D

tumorr motion in lung due to breathing and

heartbeat,, measured during radiotherapy

Yvettee Seppenwoolde, Hiroki Shirato,, Kei Kitamura, Shinichi Shimizu, Marcell van Herk, Joos V. Lebesque, Kazuo Mtyasaka

Precisee and real-time measurement of 3D tumor motion

inn lung due to breathing and heartbeat,

measuredd during radiotherapy

Inn this wortc thnee-ctimenskxiaJ (3D) motion of king tumors during radiotherapy was investigatedd in realtime . Understanding the behavior of tumor motion in tung tissue to modell tumor movement is necessary for accurate (gated or breath-hold) radiotherapy or CTT scanning. Twenty patients were included in the study. Before treatment a 2 mm gold markerr was implanted in or near the tumor. The 3D position of the irrtpianted g ^ manner wass determined by a real-time tumor tracking radiotherapy system (RTRT) using two fluoroscopyy image processor units. The system provided the coordinates of the gold markerr at a sample rate of 30 Images per second in all directions simultaneously. The recordedd tumor motion was analyzed in terms of the amplitude and curvature of the tumor motionn in three directions, the differences in breathing level during treatment hysteresis (thee difference between the inhalation and exhalation trajectory of the tumor) and the amplitudee of tumor motion induced by cardiac motion. The average amplitude of the tumor motionn was greatest (12 2 mm (SD)) in the cranial-caudal direction for tumors situated inn the lower lobes and not attached to rigid structures such as the dtest wall or vertebrae. Forr the lateral and the anterior-posterior directions, tumor motion was small, both for upperr and lower lobe tumors (2 1 mm). The tumor motion was modelled as a sinusoidal movementt with varying asymmetry. Intrafractionalry, the tumor position in the exhale phasee was more stable than the tumor position in the inhale phase. However, in many patientss shifts in the exhale tumor position wete observed intra- and interfractionaJly. The 3DD trajectory of the tumor showed hysteresis for 50% of the tumors. Fourier analysis revealedd mat for 7 of the 21 tumors, a measurable motion In the range of 1-4 mm was causedd by the cardiac beat Tumor motion due to hysteresis and heartbeat can lower treatmentt efficiency in RTRT gated treatments or lead to a geographical miss in conventionall or active breathing controfted (ABC) treatments.

Introduction n

Recentt developments in radiotherapy such as intensity modulated radiotherapy, non-coptartar corrforrrtall radiotherapy (Armstrong 1997, Robertson 1997, Beiderbos 2000, Mehta 2 0 0 1 , Haymann 2001) and active breathing controlled (ABC) gated treatments (Wong 1999), are all aimedd at increasing the tumor dose and/or reducing the dose to normal tissue. Discrepancies inn the position of a tumor between a planning computed tomography (CT) scan and during treatmentt can be caused by setup errors and organ motion (Bel 1994, Batter 1996, de Boer 2001).. To account for these errors, the clinical target volume (CTV) is expanded with a safety marginn to obtain the planning target volume (PTV). Reduction of the setup error and understandingg of organ motion are essential in designing the tightest possible safety margin withoutt c c ^ r o m i s i n g the tumor coverage. To examine the motion of lung tumors during respiration,, Ekbergg et al. (1998) used fluoroscopy at the time of simulation. They demonstrated ann average movement of 3.9 mm (range 0-12 mm) in the cranio-oaudal direction, 2.4 mm (rangee 0-5 mm) in the medio-lateral direction and 2.4 mm (range 0-5 mm) in the dorso-ventral direction.. Breath-hold (Hanley 1999, Wong 1999) or gated (Kubo 20OOa,b) radiotherapy is designedd to reduce tumor motion due to breathing. A novel method to treat moving tumors with

Chapters Chapters

aa small margin with accurate dose delivery is to use the real-time tumor tracking radiotherapy (RTRT)) system (Shirato 1999, 2000afb). Three-dimensional (3D) treatment planning is often

performedd on a CT scan made while patients breathe freely, under the assumption that the CT imagess represent the average position of the tumor. However, breathing motion can cause mis-detectionn of the tumor during CT scanning, especially for small tumors, resulting in a smallerr planning volume or a distorted tumor shape (Shtmizu 2000). To obtain an accurate tumorr image, a hold CT is preferable. The ideal breathing phase in which the breath-holdd CT has to be taken must correspond to the average tumor position. When the breathing motionn of the tumor is not symmetric, the average tumor position is no longer midway between thee in- and exhale tumor position.

Inn this study, precise 3D recordings of the tumor position were made during RTRT treatment duringg beam on and beam off period at a high sampling rate to determine and model tumor motionn due to breathing, heartbeat and patient motion.

Patientss and methods

Twentyy patiënte with tumors at different sites in the lung (one patient had two tumors) were includedd in the analysis (Table 1). Seventeen patients had non-small cell primary lung cancer andd three had metastatic lung tumors. A typical treatment schedule consisted of 4 x 10 Gy wfth aa 4 field non-coplanar conformal technique using multiple MLC-shaped static beams (4 MV).

Tablee 1. Patient characteristics.

Nrr Age Kl* Pathology y Markerr T-stage

location n

Dosee Tumor size

(Qy)) (cm8) 1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9-a a 9-b b 10 0 11 1 12 2 13 3 14 4 15 5 16 6 17 7 18 8 19 9 20 0 77 7 52 2 72 2 70 0 70 0 72 2 77 7 76 6 83 3 43 3 69 9 55 5 68 8 57 7 84 4 41 1 81 1 67 7 81 1 70 0 ** Kamofeki h 70 0 80 0 80 0 80 0 50 0 80 0 70 0 70 0 80 0 90 0 90 0 70 0 40 0 70 0 80 0 80 0 80 0 90 0 80 0 90 0 idex. . squamouss cell squamouss cell squamouss cell adenocarcinoma a squamouss ceil squamouss cell squamouss cell adenocarcinoma a squamouss cell adenocarcinoma a adenocysticc carcinoma, metastatic c squamouss cell adenocarcinoma a squamouss cell squamouss cell, metastatic

squamouss cell RCC,, metastatic adenocarcinoma a Bronchio-alveolar r adenocarcinoma a squamouss cell near r near r in n near r in n in n near r in n near r near r near r in n near r near r in n near r in n in n in n near r in n 2 2 1 1 4 4 2 2 3 3 3 3 1 1 1 1 1 1 1 1 --2 --2 2 2 2 2 --1 --1 --3 --3 3 3 2 2 1 1 40 0 35 5 40 0 40 0 40 0 40 0 40 0 40 0 40 0 40 0 40 0 35 5 35 5 40 0 35 5 40 0 40 0 35 5 35 5 40 0 40 0 14 4 85 5 22 2 20 0 10 0 36 6 60 0 16 6 10 0 14 4 8 8 92 2 3 3 135 5 23 3 14 4 168 8 257 7 2 2 9 9 25 5

3D3D tumor motion in lung

AA 2 mm gold marker was implanted into or near the tumor mass using bronchial endoscopy or byy percutaneous needle insertion if the tumor was situated near the thoracic wail. Free breathingg CT scans were taken of the entire lung volume using 5 mm slice thickness, 5 mm slicee interval. At the level of the tumor and the inserted marker the slice thickness arid the slice intervall were 1 mm or, If the tumor was large, 3 mm. Additionally, voluntary breath-hold CT scanss were made at the inhalation and exhalation phase of tidal breathing with the same slice thicknesss as for the free breathing scan. A physician delineated thé gold marker and the CTV inn each slice and the position of the marker relative to the tumor was determined. Of the three CTT scans, the scan that yielded the best possibility to irradiate the tumor (concerning normal tissuee damage, tumor coverage and treatment efficiency) was chosen as the planning CT scan.. 3D treatment planning was done using the Focus system (CMS, St Louis, MO). Thee fluoroscopic real-time tumor tracking radiotherapy system consists of four sets of diagnosticc fluoroscopy, image processor units, a trigger control unit, an image display unit and aa conventional linear accelerator with multi-leaf collimators (Shirato 2000b). Using two of the fourr fluoroscopy image processor units, the system determines the 3D position of the gold markerr 30 times a second by using real time pattern recognition and calibrated projection geometry.. To avoid blocking of the fluoroscopic images by the gantry of the linear accelerator, anyy 2 of the 4 X-ray systems may be selected. During the setup stage of each treatment fraction,fraction, Hie patient is positioned using skin markers. The physician judges both the tumor positionn and motion based on two fluoroscopy images. The position of the treatment couch is adjustedd if the gold marker is not found within the planned region. The zero tumor position is sett by the physician at the position which best corresponds to the intended treatment breathing

phase,, based on the planning CT (in most cases this was the exhale phase). Because the placementt of the zero position may slightly differ from day to day, the error made in this positionn is described as the 'residual setup error1.

Duringg treatment, the linear accelerator was triggered to irradiate the tumor only when the gold markerr was located within a specified region of the planned coordinates relative to the isocenter.. The delay between recognition of the marker and the start of irradiation was 0.09 seconds.. Radiation oncologists defined the distance of the permitted dislocation for the tumor inn the lateral <LR), cranial-caudal (CC) and ventral-dorsal (AP) directions, resulting in a box aroundd the isocenter The permitted displacement for each patient was typically 3x3x3 mm3. Thee LR, CC and AP coordinates of the internal marker were recorded 30 times per second duringg the treatment. Analog video images of the two fluoroscopic images were recorded simultaneously.. All available data sets (each fraction and each beam direction) for all patients weree analyzed; on average 16 sets of 300 seconds per patient The raw data of the RTRT systemm were filtered with a 30-point median filter to reduce system noise. For each data set thee individual breathing cycles were detected automatically by thresholding, after manually removingg inconsistent readings of the inferred position of the marker. These inconsistent readingss occurred during about 1% of the treatment time, depending on the visibility of the gold markerr on the fluoroscopic images. Different parameters were measured during the individual breathingg cycles: the amplitude, the position of the tumor in the inhale and exhale phase, the averagee tumor position and the length of the breathing cycle. Alt the individual breathing cycles recordedd during one beam direction were averaged.

Thee position s of the tumor as a function of time t can be defined as (Lujan 1999):

ChapterChapter 8

wheree s0 is the position of the tumor at exhalation, S is the amplitude, hence s0-S is the

positionn at inhalation, x is the period of the breathing cycle in seconds and <J> is the starting phase.. This parameterized breathing curve was fitted (using a least squares method) through eachh average breathing cycle. Hysteresis was assessed by the phase difference in the averagee breathing curves in two directions (Figure 1).

A,BB = amplitude x = period <j>= phase A A XX = X0-A*COS 2n (7tt/X-<|)) "2 2 A A < < B B rr V T \

JJ s

11 exhale // inhate / / / / / / f f yy = y0-B*cos2n(7d/T-(|>) B B y y X X

c c

"> > D D XX = x0-A*cos 2n (7it/x - (j>)

\\ r\

\\ \

VV \

V V

r r

i i

\\ / \\ / Acbw w yy = y0-B*cos2n(7tt/x- ((jH-Aj))) y y & &

1 1

X X

FigureFigure 1. The principle of hysteresis. If there is no phase difference between two sinusoid signals with equal frequency

(A),(A), there is no hysteresis (B); the position of the y direction is directly related to that of the x direction. When there is aa phase difference <(> (C), there are two possible positbns where the tumor is at the same x position, depending on the positionposition in the breathing cycle (D). The width and the shape of the hysteresis ellipsoid (D) depend on the phase differencedifference and the amplitude of both signals.

Thee amount of clinically relevant hysteresis was determined as the maximum distance betweenn the different trajectories that the tumor followed during inhalation and exhalation. Too separate tumor motion caused by heartbeat and breathing, the discrete Fourier transform off the unfiltered data was determined, resulting in the power spectrum of the signal. Fourier analysiss breaks down a signal into constituent sinusoids of different frequencies. The power of aa certain frequency represents the strength of this frequency in the time signal. Typical breathingg in this data set has a frequency of 0.2-0.3 Hz (a 3-5 second cycle). The heart beats withh a frequency of about 1 Hz (60 bpm). The breathing frequency and the cardiac frequency doo not overlap. When a peak of around 1 Hz was detected in the order of the power of the

3D3D tumor motion in king

breathingg signal, the raw signal was filtered with different settings (a high pass dfgital filter with aa cut-off of 0.5 Hz to eliminate the breathing frequency) to determine the amplitude of the motionn due to heartbeat in all three directions.

Results s

Amplitude Amplitude

Forr each patient the average amplitude of tumor motion in ail three directions was assessed (Tablee 2). In the cranial-caudal (y) direction, tumors that are situated in the lower lobes and are nott attached to rigid structures such as the chest wall or vertebrae move more than upper lobe tumorss or tumors attached to rigid structures: 12 6 and 2 2 mm (SD), respectively, (p=0.005,, two tailed, unequal variances).

Tablee 2. The mean amplitude of tumor motion, related to tumor location and attachment, the value off n for the fitted breathing curve and the length of the breathing period.

Meann amplitude* Breathing period (mmm ISP) ( s t ISP) Nrr LR CC AP Lobe Tumor attached to n x 20 0 9~a a 10 0 3 3 7 7 18 8 17 7 16 6 19 9 6 6 8 8 12 2 9-b b 1 1 5 5 11 1 15 5 14 4 4 4 2 2 13 3 1.88 (0.4) 1.00 (0.3) 1.44 (0.3) 0.66 (0.4) 2.55 (0.3) 2.00 (0.4) 0.44 (0.1) 0.55 (0.1) 1.11 (0.3) 0.22 (0.1) 0.66 (0.1) 1.33 (0.2) 2.88 (0.6) 2.88 (0.3) 0.44 (0.1) 1.77 (0.3) 0.77 (0.1) 0.66 (0.1) 0.66 (0.1) 2.44 (0.4) 0.66 (0.1) ZiAZiA (3.8) 13.11 (2.7) 11.88 (0.9) 14L££ (0.8) L7__ (0.9) MM (1.0) 1.00 (0.3) 0.22 (0.1) 11.11 (2) 0.77 (0.1) M.M. (0.8) 4.22 (0.6) 3.00 (0.5) 2.88 (0.4) 2.88 (0.6) 2.66 (0.4) 2.00 (0.4) 1.66 (0.2) 1.22 (0.2) 1.00 (0.2) 0.77 (0.1) 1.9(0.4) ) 3.6(0.5) ) 2.6(0.5) ) 1.7(0.5) ) 0.9(0.1) ) 0.9(0.2) ) 1.7(0.5) ) 0.2(0.1) ) &8<0.7) ) 1.2(0.1) ) &2(0.5) ) 4.11 (0.5) 2.6(0.5) ) 1.6(0.2) ) 0.8(0.3) ) 1.4(0.2) ) 2.7(0.2) ) 1.4(0.2) ) 2.0(0.4) ) 1.2(0.1) ) 0.6(0.1) ) Lower r Lower r Lower r Lower r Lower r Lower r Lower r Lower r Middle e Middle e Upper r Upper r Upper r Upper r Upper r Upper r Upper r Upper r Upper r Upper r Upper r aorta a

anteriorr chest wall posteriorr chest wall

laterall chest wall

vertebra,, aortic arch

bronchial l aorta a

laterall chest wall aorta a aorticc arch 1 1 2 2 1-2 2 2 2 2 2 3-9 9 1 1 1 1 2 2 2 2 2-3 3 2 2 1 1 2 2 2 2 1 1 1 1 1-2 2 1 1 1 1 1 1 3.88 (0.6) 3,66 (0.4) 4.00 (0.3) 3.00 (0.3) 3.33 (0,8) 6.66 (1.5) 3.88 (0.6) 4.99 (0.8) 3.22 (0.3) 4.88 (0.5) 3.88 (0.3) 3.22 (0.4) 3.55 (0.3) 2.88 (0.2) 2.77 (0.4) 3.22 (0.4) 3.22 (0.3) 3.77 (0.4) 3.22 (0.4) 3.33 (0.3) 3.77 (0.5) ** Tumor motion greater than 5 mm is underlined.

ChapterChapter 8

Forr tumors attached to rigid structures and for the LR and AP direction there was no difference inn amplitude between upper and lower lobe tumors; the motion was small in these directions: 1.22 0.9 mm (SD) for the LR, and 2.2 1.9 mm (SD) for the AP direction. Two patients showed tumorr motion of more than 5 mm in the AP direction, both tumors were located anterior and in thee middle of the thorax (patient 8 and 19). The average efficiency (total treatment time/total irradiationn time) of the RTRT system is 59% for this patient group. The efficiency depends on thee amplitude of the breathing motion and the size of the box representing the permitted displacement.. If the tumor motion was large, the relative amount of time that the tumor spent outsidee the permitted displacement region was large and the treatment efficiency low. If the treatmentt efficiency dropped below the 10%, the RTRT treatment was abandoned and the patientt was treated using a conventional treatment technique.

Averagee tumor position

Sixx patients showed a time-trend in tumor position during one beam direction or during one treatmentt day, mostly in the posterior direction. Between the beams, shifts in the tumor positionn were frequently observed in all 3 directions (for example patient 3, Figure 2A).

I I nhalation n exhalation n -20 0

l 4 ^ ^ 4 i / HH >VVv^^4Hw4

nhalation n exhalation n a

V * V ^ ^

((Xhalation n ihhalation n 2000 300 400 nrr of breathing cycle 500 0 600 0

FigureFigure 2A. Position of in- and exhalation phase of patient 3 during the treatment. I. LR-direction. II. CC-direction. III.

AP-direction.AP-direction. The numbers represent the days and the beam directions, for example, 21 is the first beam on day 2. The variationvariation in the inhale positbn is larger than in the exhale position. Between days and between beams a systematic shiftshift in the average tumor positbn can be present (thick arrow). In the AP and LR direction a gradual shift is present in thethe course of one treatment day (thin arrows). (The treatment efficiency of beam 15 and beam 31 was too low and this beambeam was delivered without using the RTRT system with a larger margin).

3D3D tumor motion in lung Dayy 1

77 <[vArfr..

<Mt t

Dayy n > > 4 4 >1.2 2 Beamm 1 2 Beamm 1 2

F/guree 2B. 77?e different variations in tumor position that are given in Table 3 (the numbers in the figure correspond to

thethe numbers used in Table 3):

1.1. The average tumor position represents the difference between the planned zero-position and the actual tumor positionposition during the treatment averaged over all patients. The systematic deviation from 0 is because the planned zero-positionposition was normally the exhale and not the average tumor position.

2.2. The variation in the average tumor position contains variation in residual setup errors, breathing, patient movement duringduring and between treatment days, changes in breathing level and intensity.

3.3. The variation in the average inhale tumor position contains variation in residual setup errors, breathing, patient movementmovement during and between treatment days, changes in breathing level and intensity.

4.4. The variation in the average exhale tumor position, variation in residual setup errors, breathing, patient movement duringduring and between treatment days and changes in breathing level.

6.6. The variation in the inhale tumor position, corrected for its average, contains variation in patient movement during treatment,treatment, changes in breathing level and intensity.

7.7. The variation in the exhale tumor position, corrected for its average, contains variation in patient movement during treatmenttreatment and changes in breathing level.

Tablee 3. Summary statistics of tumor positions (in mm).

LR R CC C AP P

1.. Average tumor position 2.. SD in average tumor position 3.. SD in average inhale tumor position 4.. SD in average exhale tumor position 5.. SD in amplitude

6.. SD in inhale tumor position* 7.. SD in exhale tumor position*

-0.4 4 0.8 8 0.9 9 0.8 8 0.2 2 0.2 2 0.2 2 -1.5 5 1.2 2 1.3 3 1.2 2 0.8 8 0.8 8 0.4 4 -1.6 6 1.1 1 1.1 1 1.1 1 0.3 3 0.4 4 0.3 3 'Correctedd for changes in average exhale or inhale position.

Betweenn subsequent days, the average tumor position could shift as well. The trends and the shiftss during one treatment day may have been caused by patient motion or internal tumor motionn (change in breathing level) while for the shifts between subsequent days the 'residual setupp error' would have played a role as well. The average tumor position and the variations inn the average tumor position, in the amplitude and in the inhale and exhale tumor positions, averagedd for each treatment beam, or corrected for its average (Figure 2B), are given in Table 3.. The variation in the averages of the exhale and inhale tumor position were comparable to thee corresponding values of the average tumor position.

Forr the CC and the AP direction, the variation in the inhale and exhale tumor position, correctedd for the average values (and thus corrected for the shifts between the beams and betweenn the days) was smaller (p-value < 0.001, paired t-test) than the variation in the inhale position. .

ChapterChapter 8

3DD path of the tumor

Visualizationn of the 3D path of the tumor (Figure 3A) gives an impression of the hysteresis and thee relation of the tumor motion to the permitted displacement (gate range). The shape of the averagee 3D curve (Figure 3B) was fairly constant in time; differences were mainly due to shifts inn average tumor position and changes in amplitude. The projections of the trajectories of the tumorss on the coronal and sagittal plane were calculated for each patient (Figure 4). A change inn the shape of the 3D path was observed between subsequent days in only one patient (Figuree 5).

Gatee range

LRR (mm)

FigureFigure 3. A. The 3D path of the tumor of patient 10 during one treatment portal, the gray dots represent the tumor

positionposition throughout the treatment, the black dots represent the tumor position as the tumor is detected to be inside the RTRT-rangeRTRT-range (transparent box). B. The average 3D curve of the tumor. The projections on the coronal, sagittal and axialaxial plane are drawn in thin black lines. Note that the tumor follows a different path during inhalation than during

exhalationexhalation (hysteresis).

FigureFigure 4. Orthogonal projections of the trajectories of the 21 tumors on the coronal and the sagittal plane. The tumors

areare displayed at the approximate position, based on the localization mentioned in the treatment chart. Tumors that were attachedattached to bony structures are colored red, lower lobe tumors are colored light blue.

ChapterChapter 8

3DD path of the tumor

Visualizationn of the 3D path of the tumor (Figure 3A) gives an impression of the hysteresis and thee relation of the tumor motion to the permitted displacement (gate range). The shape of the averagee 3D curve (Figure 3B) was fairly constant in time; differences were mainly due to shifts inn average tumor position and changes in amplitude. The projections of the trajectories of the tumorss on the coronal and sagittal plane were calculated for each patient (Figure 4). A change inn the shape of the 3D path was observed between subsequent days in only one patient (Figuree 5).

Gatee range

LRR (mm)

FigureFigure 3. A. The 3D path of the tumor of patient 10 during one treatment portal, the gray dots represent the tumor

positionposition throughout the treatment, the black dots represent the tumor position as the tumor is detected to be inside the RTRT-rangeRTRT-range (transparent box). B. The average 3D curve of the tumor. The projections on the coronal, sagittal and axialaxial plane are drawn in thin black lines. Note that the tumor follows a different path during inhalation than during

exhalationexhalation (hysteresis).

FigureFigure 4. Orthogonal projections of the trajectories of the 21 tumors on the coronal and the sagittal plane. The tumors

areare displayed at the approximate position, based on the localization mentioned in the treatment chart. Tumors that were attachedattached to bony structures are colored red, tower lobe tumors are colored light blue.

3DD tumor motion in lung

Shapee and length of the breathing cycle

Forr the patient group in this study, the average length of one breathing cycle was 3.6 0.8 secondss (SD). The individual breathing periods and the variation during the treatment are givenn in Table 2. For each patient, the parameters of the sinusoid (Equation 1) were determined.. Equation 1 fitted well for the majority of patients (Figure 6A), the fitted values for thee amplitude and period agreed with the measured values. The value for n varied between thee patients (Table 2).

Exhale e

CC(mm) )

LR(mm) ) CC(mm) ) LRR (mm)

FigureFigure 5. The 3D trajectory of the tumor of patient 8. A. On day 1, the pattern was similar to that on day 2 and 4.

B.B. On day 3. The shape and direction of the motion changed between subsequent days: in the coronal plane on day 3 hysteresishysteresis was present, while on days 1,2 and 4 this was not the case. In the transaxial (AP-LR) plane the motion changedchanged from left to right on day 1,2 and 4 to right to left on day 3.

x=6.6s s Patientt 18

x=10.11 s Tbne(s) )

FigureFigure 6. A. The time signal of the tumor motion of patient 7 (CC direction). A curve was fitted through the data with

tt = 3.3sandn = 2. B.The time signal of the tumor motion of patient 18 who had a very irregular breathing pattern (CC(CC direction). The thin Tine represents tumor motion with a period of 6.6 s, while later on the period was 10.1 s. The firstfirst cycle could be Med well with n = 3,the second cycle was fitted withn = 9. Note that the shape of the tumor motion ofof the longer period is similar to that of the shorter cycle; only the time spent in the exhale position is prolonged in the longerlonger cycle.

ChapterChapter 8

Forr two patients the parameters n=1 and n=2 fitted equally well, suggesting that the shape of thee actual breathing curve was somewhere between the two parameterized curves. In one case,, the patient's breathing was very irregular and the value for n varied between 3 and 9 (Figuree 6B), depending on the breathing period; the longer the period, the higher the value off n. However, the shape of the tumor motion for the longer period was similar to that of the shorterr cycle; only the time spent in the exhale position was prolonged in the longer cycle. The valuee of n did not correlate with the breathing period for the whole patient group.

Hysteresis s

Inn the (average) trajectories of the tumor, hysteresis was observed in 10 of the 20 patients: the tumorr followed a different path during inhalation than during exhalation. The presence of hysteresiss in one plane could be determined by calculating the phase difference between the fittedd parameterized curves of the average breathing cycles of two directions (Figure 7). When present,, hysteresis was largest in the AP-CC plane (sagittal plane) but was also observed in thee other planes (Table 4). A phase shift between the signals did not lead to a measurable hysteresiss in 4 more patients because the tumor movement was less than 1 mm.

A.. Patient 8

Timee (s)

FigureFigure 7. A. Time signal of patient 8 in the AP and the CC direction. There was a phase difference (A$) of 0.5 s between

bothboth signals, causing hysteresis. At the arrow the tumor in the CC direction is still in the exhale position while in the AP directiondirection inhalation already started. B. Time signal of patient 18 in the AP and the CC direction. There was no measurablemeasurable phase difference between both signals.

3DD tumor metion in kmg

Tablee 4. The amount of hysteresis (in mm) in three ptanes.

Nr r Coronall (mm) Sagittal (mm) Transaxial (mm) 1 1 2 2 8 8 9-a a 9-b b 10 0 11 1 12 2 15 5 19 9 20 0 1.5 5 15 5 5 5 2 2 2 2 1 1 1 1 1 1 2 2 3 3 1 1 2 2 2.5 5 3 3 1.5 5 Heartbeat t

Inn seven of the 20 patients, the heartbeat caused a measurable tumor motion (Table 5). If the cardiacc frequency was detected in the frequency analysts (Figure 8A), the amplitude of the tumorr motion was measured in the time signal (Figure 8B). In two patients tumor motion due too the heart was detected in two directions. The tumor was attached to, or near the aorta in six off the seven patients. The distance of the marker to the cardiac or aortic wall was less than 3 cmm in these six patients. The motion due to the heartbeat was largest in the LR direction; range 1-44 mm. In the other directions the motion was 1-2 mm.

Tablee S. Tumor motion due to heartbeat

Nr r 2 2 5 5 7 7 8 8 9-b b 11 1 15 5 Frequency y (bpm) ) 66 6 70 0 62 2 60 0 63 3 60 0 72 2 LR R 1 1 1 1 2 2 4 4 Amplitude e (mm) ) CC C 1 1 2 2 1.5 5 AP P 1.5 5 1 1 Lobe e upper r upper r lower r upper r upper r upper r upper r Tumor r attached d to o aorta a bronchus s aorta a free free vertebraa & aorticc arch aorta a free e Distance1 1 (mm) ) 30* * 10 0 25* * 30 0 15* * 27** * 65* * 'fromm the cardiac or aortic wait to the marker.

Thee discrepancies between the tumor attachment and the distance from the cardiac or the aortic walll to the marker can be explained by: * the marker was positioned near and not in the tumor; *** the tumor volume was large.

ChapterChapter 8

I I

CL L 44

-rW....

J

breathingg cycle 55 s .. . ^ __. 11 heartbeat LL m

J J

AP P 0.5 5 1.5 5 44 -2 -2

QtflA^A. .

breathingg cycle s s uA. . 0.55 1 Frequencyy (Hz) CC C 1.5 5

FigureFigure 8. A. The frequency spectrum of the LR and CC coordinates of patient 11. Two frequency peaks are present in

thethe registration, one at 0.3 Hz due to breathing and one at 1.05 Hz due to the heartbeat. In the CC direction the power ofof the heartbeat signal is weak compared to the power of the breathing signal, in the LR direction, the heartbeat has thethe largest power.

CC,, amplitude 3 mm (breathing) AP,, amplitude 4 mm (heart beat)

Timee (s)

FigureFigure 8. B. The time signals of the LR (solid line) and CC (dotted line) tumor positions. In the LR direction tumor motion

3D3D tumor motion in king

Discussion n

Thee RTRT system is unique in recording the tumor position in all three directions simultaneouslyy at a high sampling rate. This enabled us to detect tumor motion due to the heartbeat,, as well as hysteresis. The system measures the position of a gold marker implanted inn or near the tumor. In some studies (Hartley 1999, Man 2000, Minohara 2000, Kubo 2000b) thee position of the chest waft or diaphragm is used to monitor breath-holding or to trigger the linearr accelerator. However, the position of the tumor can be different to the position of these structures.. For example, in the study of Hanley et al. (1999) the chest wall moved with an amplitudee of 2-2.5 mm while the diaphragm of the same patients moved 20-38 mm. Minohara ett al. (2000) measured the breathing phase with an tr-LED and a PSD camera. They found a phasee difference of about 200 ms between the position of the diaphragm and the respiratory Signall from the LED-camera system which was placed under the left rib. In other studies the volumee of the inhaled air is used as a measure of the tumor position (Wong 1999). While these techniquess are useful to determine the in- or exhale phase of the breaming, the 3D tumor motionn and the exact position of the tumor are not measured directly and must be inferred. Duringg real time tumor tracking radiotherapy, the shape of the 3D path of the tumor did not changee significantly over time. However, its amplitude and position in space varied due to to patentt shift or 'setup' errors as well as changes in breathing level and intensity. The tumor motionn due to breathing is not one-dimensional, as assumed in simulation studies {Lujan 1999, Engelsmann 2001a) but a combination of movement in all three dimensions, sometimes resultingg in hysteresis. However, tumor motion due to breathing is largest in the cranial-caudal direction,, especially in unfixed tower lobe tumors. Tumor motion in the presented patient group agreess with tumor motion found by Ekberg et al. (1998). In the present study, the amplitude of thee movement was different for tumors attached to rigid structures such as the chest wall or vertebrae.. These tumors only move slightly, in agreement with findings of Ross et al. (1990) whoo also found that tumors in upper lobes and tumors that were attached to the chest wall movedd less than lower lobe tumors.

Twoo patients (patents 8 and 19) showed significant hysteresis. Bom tumors were located at a similarr position in the lungs, one in the left and one in the right lung (Figure 4). One of these patiëntee suffered from emphysema with prominent chest breathing. These findings suggest thatt pre-existing lungg disease and breathing technique are important for estimating hysteresis inn the tumor motion.

Thee position of the tumor during the exhale phase of tidal breathing is considered to be the mostt stable and reproducible position (Baiter 1998). This is assumed in many gated treatment designss and for the intrafractional period this was confirmed by our study. However, for some tumors,, a systematic shift or trend between different beams and/or fractions was observed, particularlyy in the posterior direction. The variations in the shifts and trends were larger than thee variations during one beam and do not differ for the inhale and exhale tumor position. Thee time trends could be attributed to patient relaxation throughout the treatment (combined withh a reduction in amplitude) or to gravity acting on compliant lung tissue shortly after the patientt assumes a supine position. A systematic shift could be due to changes in muscle tone, aa slight shift in patient position or simply because the patient started breathing at a different level.. The shifts in exhale position between subsequent days can also be caused by the variationn in the zero-position of the RTRT treatment. During the setup phase of the treatment thee physician assesses the tumor motion based on two fluoroscopy images. The tumor motion visiblee on these images is a 2D projection of the real 3D motion. It is possible that the

zero-Chapters zero-Chapters

positionn along one direction is chosen correctly, but not necessarily for the other directions. Additionally,, it is not possible to put the zero-position at exactly the same position in the breathingg cycle each day. Thus, the combination of the placement of the zero-position, patient shiftss and changes in breathing level can cause a variation in the order of 1-1.5 mm. Improvementt in the placement of the zero-position is required for Improved accuracy. This can bee accomplished by automatically determining the average tumor position relative to the exhalee tumor position by means of a short real-time tumor tracking session before the actual treatment.. However, inter-fraction changes in breathing level cannot be monitored with this technique. .

Too obtain the best treatment efficiency for RTRT treatments as well as to determine the ideal tumorr position for gated or breath-hold radiotherapy or CT, the phase of the breathing cycle representativee of the average tumor position must be identified. Tumor motion can be modeled usingg a periodic but asymmetric function (more time spent at exhale versus inhale; Equation 1).. The amount of time the tumor spent in the exhale position differed per patient. Tumor motionn was more asymmetric for a patient with a relatively long breathing cycle.

Thee 3D analysis of the tumor motion also revealed that in some patients the trajectory of the tumorr during inhalation is different from the exhalation trajectory. This hysteresis can, for example,, be induced by the breathing technique, especially when diaphragm and chest breamingg is combined asymmetrically. Another possible explanation for hysteresis is the dynamicc properties of the lung. Due to the elasticity of lung tissue, tumor motion may be delayedd compared to the motion of the chest wall and/or the diaphragm. It can take a while beforee the tumor occupies a stationary position (this is reached during the relatively long Vest* periodss during the inhalation and especially the exhalation phase). Lung diseases such as emphysema,, bullae or fibrosis can influence the elasticity of lung tissue anisotropically and causee hysteresis. When hysteresis in tumor motion is caused by the dynamic properties of lungg tissue, breath-hold scans will not give the representative position of the tumor. During normall breathing there is hysteresis, however, during a long breath hold, the tumor occupies itss stationary position. Hysteresis can seriously affect the accuracy in radiotherapy, which uses thee position of skin, diaphragm, or physiological parameters for respiration gating without consideringg the phase difference with respiration. During real time tumor tracking the tumor mayy be inside the permitted range for inhalation, but outside for exhalation (Figure 9). If this occurss in two or three directions, hysteresis can increase the irradiation time considerably. The causee of hysteresis and its effect on treatment accuracy will require further study in the future. Usingg discrete Fourier analysis, tumor motion due to heartbeat was detected in 7 of the 20 patients.. The amplitude of this motion was 1-4 mm, mostly in the LR direction. Tumor movementt with heartbeat was most significant for tumors attached to the aorta. In the fluoroscopyy study of Ekberg et al. (1998) it was observed that for tumors located close to the heartt cardiac movement was a major contributor to the tumor motion. In a 20-patient study of Rosss et al. (1990) using an ultra fast CT-scanner, tumor motion of 9 6 mm in the lateral and APP direction was measured which was attributed to aortic pulsation, cardiac contraction and respiration.. Neither author distinguished between tumor motion caused by breathing or cardiac motion. .

Althoughh real-time tumor tracking using an implanted gold marker resulted in precise informationn regarding the tumor position, the technique has inherent limitations. The gold markerr is not always inserted exactly into the tumor; so, the motion measured with the RTRT systemm may not correspond exactly with the real tumor motion. Measuring marker motion relativee to tumor motion during treatment is hampered because of the poor visibility of tile

3D3D tumor motion in lung

tumorr in the fluoroscopic images. If the tumor is large, some parts of tumor may not move as aa fixed rigid structure. Tumor rotation and deformation cannot be detected using a single insertedd marker. Furthermore, the presence of the gold marker (relatively heavy compared to thee lighter lung tissue) may distort the tumor motion or result in hysteresis. Migration of the internall marker was not studied in lung cancer, but was small (in the same order of the CT accuracy)) in prostate and liver cancer studies (Kitamura 2000). Experiments involving implantationn of 3 markers to measure rotation and tumor shrinkage have already begun. Breath-holdd and repeat CTs will be made to assess marker migration and tumor motion relativee to the marker motion.

Pathh of the tumor withh hysteresis

V... .

_{• •} • • * * \ \ •» » ^ ^^ Linac on Allowed d tumorr position a a

i i

* *

N N

Ü Ü

• • . . . ^^ k . 3 3 m n T T \ \ • • \ \ f f

w w

T T

i i

Pathh of the tumor withoutt hysteresis

FigureFigure 9. The path of a moving tumor with and without hysteresis (dotted line) during RTRT treatment. The solid line

representsrepresents the time that the linear accelerator is irradiating the tumor. If hysteresis is present the tumor is only irradiated duringduring about 10% of the treatment time, which number increases to 30% if there is no hysteresis.

Conclusion n

Thee RTRT system has been used to measure the tumor position in all three orthogonal directionss simultaneously, at a high sampling rate that enabled the detection of tumor motion duee to heartbeat as well as hysteresis. Tumor motion and hysteresis could be modelled with ann asymmetric trigonometric function. Tumor motion due to breathing was greatest in the cranial-caudall direction for lower lobe, unfixed tumors.