Respiration tracking in radiosurgeryAchim SchweikardInformatik, Universitaet Lu¨beck, 23538 Lu¨beck, Germany

Hiroya ShiomiDivision of Multidisciplinary Radiotherapy, Osaka University Graduate School of Medicine, Japan

John AdlerNeurosurgery, Stanford University Medical Center, Stanford, California 94305

~Received 21 July 2003; revised 26 May 2004; accepted for publication 1 June 2004!

Respiratory motion is difficult to compensate for with conventional radiotherapy systems. An ac-curate tracking method for following the motion of the tumor is of considerable clinical relevance.We investigate methods to compensate for respiratory motion using robotic radiosurgery. In thissystem the therapeutic beam is moved by a robotic arm, and follows the moving target through acombination of infrared tracking and synchronized x-ray imaging. Infrared emitters are used torecord the motion of the patient’s skin surface. The position of internal gold fiducials is computedrepeatedly during treatment, via x-ray image processing. We correlate the motion between externaland internal markers. From this correlation model we infer the placement of the internal targetduring time intervals where no x-ray images are taken. Fifteen patients with lung tumors haverecently been treated with a fully integrated system implementing this new method. The clinicaltrials confirm our hypothesis that internal motion and external motion are indeed correlated. In apreliminar study we have extended our work to tracking without implanted fiducials, based onalgorithms for computing deformation motions and digitally reconstructed radiographs. ©2004American Association of Physicists in Medicine.@DOI: 10.1118/1.1774132#

Key words: whole body radiosurgery, respiration tracking, soft-tissue navigation

uimb-

toue

touenll

ars

da

ren-h

aolthoo

mry

th

vel-rm.

ess-

tion1 cmgonom-re-vedeion-ldsbut

el-f

rale-a

naltly

I. INTRODUCTION

The success of radiosurgical methods for brain tumors sgests that higher precision treatment could dramaticallyprove treatment outcome with tumor irradiation, and enatissue ablation.1,2 High precision radiosurgery has been limited to brain tumors, since stereotactic fixation is difficultapply to tumors in the chest or abdomen. To apply radiosgical methods to tumors in the chest and abdomen, it is nessary to take into account respiratory motion. Respiramotion can move the tumor by more than 1 cm. Withocompensation for respiratory motion, it is necessary tolarge the target volume with a ‘‘safety margin.’’ For smatargets, an appropriate safety margin produces a very lincrease in treated volume. For a spherical tumor of radiucm, a safety margin of 0.5 to 1 cm would have to be addeensure the tumor remains within the treated volume attimes. The ratio between the radius and volume of a sphecubic. Thus a margin of 1 cm will cause an eight-fold icrease in treated volume, which primarily involves healttissue. Furthermore, observed motion ranges~in excess of 3cm! suggest that a 1 cm margin may not be sufficient incases. An enlarged margin of 2 cm would result in a 27-fincrease of dose in this example. Thus, an accurate mecapable of compensating for respiratory motion would beutmost clinical relevance. Typical variations in the patternrespiratory motion can range from 1 to 2 cm for the sapatient~in pediatrics!, and the duration of a single respiratocycle has been reported to vary between 2 and 5 s.3

We have investigated the design of a new sensor me

2738 Med. Phys. 31 „…, October 2004 0094-2405

g--

le

r-c-ryt-

ge1

tollis

y

lldodffe

od

for tracking respiratory motion.4 A method for correlatinginternal motion to external surface motion has been deoped. The treatment beam itself is moved by a robotic aThe system is based on a modified Cyberknife~Accuray Inc.,Sunnyvale, CA!.

A number of patients have recently been treated succfully with this new correlation-based method.

II. RELATED WORK

Intratreatment displacements of a target due to respirahave been reported to exceed 3 cm in the abdomen, andfor the prostate.5,3 Webb6 discusses methods for achievinimproved dose conformality with robot-based radiatitherapy. Specifically, trade-offs between treatment path cplexity and conformality are analyzed. The experienceported suggests that very high conformality can be achiewith robotic systems by modulating the intensity of thbeam. However, this assumes the ideal situation of a statary target, unaffected by respiration. This assumption hoonly for few anatomic sites, such as the brain or spine,not for tumors close to the lung or diaphragm.

Conventional radiation therapy with medical linear accerators~LINAC systems! uses a gantry with two axes orotation movable under computer control.7 This mechanicalconstruction was designed to deliver radiation from sevedifferent directions during a single treatment. It was not dsigned to track respiratory motion. Respiratory gating istechnique for addressing this problem with conventioLINAC radiation therapy. Gating techniques do not direc

Õ2004Õ31„8…Õ1Õ4Õ$22.00 © 2004 Am. Assoc. Phys. Med.

eaheneacutrsetth

ti

hetinchwct

r iatthm

atkraheesoy

acfrau

ric

rn

e

u

bc

ac

po

fpsze

sndtwo

theed

ofiseonstionce,rlyal10

ghtrol

tioned to

urce

ernal

2739 Schweikardetal. : Respiration tracking in radiosurgery 2739

compensate for breathing motion, i.e., the therapeutic bis not moved during activation. Instead the beam is switcoff whenever the target is outside a predefined window. Oof the disadvantages of gating is the increase in treatmtime. A second problem is the inherent inaccuracy of suchapproach. One must ensure that the beam activation cycan have sufficient length for obtaining a stable therapebeam. Kubo and Hill8 compare various external senso~breath temperature sensor, strain gauge, and spirom!with respect to their suitability for respiratory gating wiLINACS. By measuring breath temperature, it is possibledetermine whether the patient is inhaling or exhaling. Itverified in Ref. 8 that frequent activation/deactivation of tlinear accelerator does not substantially affect the resuldose distribution. However, the application of such a tenique still requires a substantial safety margin for the folloing reason: The sensor method only yields relative displaments during treatment, but does not report and updateexact absolute position of the target during treatment.

Tadaet al.9 report using an external laser range sensoconnection with a LINAC-based system for respiratory ging. This device is used to switch the beam off wheneversensor reports that the respiratory cycle is close to maxiinhalation or maximal exhalation.

We investigate a method for tracking a tumor during trement. Stereo x-ray imaging is combined with infrared tracing. X-ray imaging is used as an internal sensor, while infred tracking provides simultaneous information on tmotion of the patient surface. While x-ray imaging givaccurate information on the internal target location, it is npossible to obtain real-time motion information from x-raimaging alone. In contrast, the motion of the patient surfcan be tracked with commercially available high speed inred position sensors. The main idea of our approach is toa series of images from both sensors~infrared and x ray!where signal acquisition is synchronized. From such a seof sensor readings and corresponding time-stamps, wedetermine a motion pattern. This pattern correlates extemotion to internal motion.

Below we describe an integrated system using the ncorrelation method~internal versus external fiducials!, andoutline a concept for extending this work to tracking withoimplanted fiducials.

III. SYSTEM

A. Overview

Figures 1 and 2 show the system components. A roarm ~modified Cyberknife system! moves the therapeutibeam generator~medical linear accelerator!. The componentadded to the standard Cyberknife system is an infrared tring system~BGI Inc., Boulder, CO!. Infrared emitters areattached to the chest and the abdominal surface of thetient. The infrared tracking system records the motionthese emitters. A stereo x-ray camera system~x-ray cameraswith nearly orthogonal visual axes! records the position ointernal gold markers, injected through an 18 gauge bioneedle into the vicinity of the target area under computeri

Medical Physics, Vol. 31, No. 10, October 2004

mdentn

lesic

er

os

g--e-he

n-eal

---

t

e-se

esanal

w

t

ot

k-

a-f

yd

tomographic~CT! monitoring. In addition, hardware meanfor capturing the time points of x-ray image acquisition ainfrared sensor acquisition are used to synchronize thesensors.

Prior to treatment, small gold markers are placed invicinity of the target organ. Stereo x-ray imaging is usduring treatment to determine the precise spatial locationthese gold markers. Using stereo x-ray imaging, precmarker positions are established once every 10 s. Limitation patient radiation exposure and x-ray generator activatimes currently prevent the use of shorter intervals. Notihowever, that even much shorter time intervals would cleabe insufficient for capturing respiratory motion. The typictarget velocity under normal breathing is between 5 andmm/s.

External markers~placed on the patient’s skin! can betracked automatically with optical methods at very hispeed. Updated positions can be transmitted to the con

FIG. 1. System overview. Infrared tracking is used to record external moof the patient’s abdominal and chest surface. Stereo x-ray imaging is usrecord the three-dimensional position of internal markers~gold fiducials! atfixed time intervals during treatment. A robotic arm moves the beam soto actively compensate for respiratory motion.

FIG. 2. Infrared emitters, attached to the patient’s chest, are used as extmarkers~Osaka University Medical Center!.

rnc

tioicenlyc

uferk,

ivthurnam

edn

avethetheial

se.on

-sig-ofhetual

innd-are

ionthtion.of

ing-

it-tershisder-

ra-

es-

on-outasefi-

ing

ingill

r

2740 Schweikardetal. : Respiration tracking in radiosurgery 2740

computer more than 20 times per second. However, extemarkers alone cannot adequately reflect internal displaments caused by breathing motion. Large external momay occur together with very small internal motion, and vversa. We have observed 2 mm external emitter excursiocombination with 20 mm internal target excursion. Similarthe target may move much slower than the skin surfaSince neither internal nor external markers alone are scient for accurate tracking, x-ray imaging is synchronizwith optical tracking of external markers. The external maers are small active infrared emitters~Flashpoint FP 5000BGI Inc., Boulder, CO! attached to the patient’s skin~Fig. 2!.The individual markers are allowed to change their relatplacement. The first step during treatment is to computeexact relationship between internal and external motion,ing a series of x-ray snapshots showing external and intemarkers simultaneously. Each snapshot has a time-stwhich is used to compute the correlation model.

IV. EXPERIMENTS AND CLINICAL TRIALS

Clinical trials of the method in Sec. III have been carriout at several institutions, including Cleveland Clinic Fou

FIG. 3. Total target excursion~top curve! and correlation error~bottomcurve! in mm for a clinical case.x-axis: treatment beam direction numbe~x-ray live shot number!, y-axis: error in mm.

Medical Physics, Vol. 31, No. 10, Otober 2004

ale-n

in,e.fi-d-

ee

s-alp

-

dation and Osaka University Hospital. Sixteen patients hsince been treated at Osaka University Hospital withfiducial-based method described in Sec. III. Notice thatmethod in Sec. III still requires the use of implanted fiducmarkers.

A. Clinical trials

Figure 3 shows representative results for one clinical caThe figure shows the total correlation error. Thus, basedthe correlation model~Sec. III!, we compute the current position of the target based on the external infrared sensornal alone. At this same time point, we also acquire a pairx-ray images. We then plot the distance in mm from tplacement inferred by the correlation model and the acplacement determined from the implanted fiducial markersthe image. The top curve in this figure shows the correspoing target excursion. Two more representative casesshown in Figs. 4 and 5.

V. DISCUSSION AND CONCLUSIONS

The above described technique combines informatfrom two sensors: x-ray imaging and infrared tracking. Bosensors have characteristic advantages in our applicaStereo x-ray imaging can determine the exact locationfiducial markers via automatic image analysis. Our trackmethod~x-ray imaging with time-stamps, combined with infrared sensing! does not require the location of infrared emters be computed in the x-ray images. The infrared emitdo not even have to be visible in the x-ray images. Tavoids occlusions and simplifies image processing consiably. The observed motion of external infrared emitters~4 to20 mm! is large when compared to the noise range of infred position computation~below 0.01 mm!. Thus, the ratiobetween noise and signal is adequate for following the rpiratory cycle.

Clinical testing of the above described method demstrates an overall correlation error of below 2 mm throughan entire 70 min treatment, while the total target motion wover 10 mm. This suggests that our method meets the dnition of radiosurgcial accuracy and is capable of reducsafety margins by a very substantial amount.

Preliminary results obtained with an extended trackmethod, which obviates the need for implanted fiducials, w

e-t

0

FIG. 4. Treatment of a hepato-cellular carcinoma~HCC!with fiducial-based respiration compensation as dscribed in Sec. III~left image: before treatment, righimage 3 months after treatment!. Total treatment timewas 40 min per fraction for three fractions with 8beam directions and 39 Gy.2

al

2741 Schweikardetal. : Respiration tracking in radiosurgery 2741

FIG. 5. Treatment of an adeno carcinoma with fiducibased respiration tracking~Osaka University Hospital!~treatment with 39 Gy/3 fractions!.

ienahoth

a-

oo

c

r,io-tion

a-ed.

y,’’

: a

h

be described in a forthcoming article. This new methodvery closely related to the method above, but adds sevelements: computation of 3D deformation images for interorgans, computation of digitally reconstructed radiograpand a technique termed 7D registration, which aims at bregistering the position and orientation of an organ, andcorresponding respiration stage.

1A. Schweikard, M. Bodduluri, and J. R. Adler, ‘‘Planning for camerguided robotic radiosurgery,’’ IEEE Trans. Rob. Autom.14~6!, 951–962~1998!.

2K. R. Winston and W. Lutz, ‘‘Linear accelerator as a neurosurgical tfor stereotactic radiosurgery,’’ Neurosurgery22, 454–464~1988!.

3M. R. Sontaget al., ‘‘Characterization of respiratory motion for pedriaticonformal 3D therapy,’’ Med. Phys.23, 1082~1996!.

Medical Physics, Vol. 31, No. 10, Oktober 2004

srall

s,the

l

4A. Schweikard, G. Glosser, M. Bodduluri, M. Murphy, and J. R. Adle‘‘Robotic motion compensation for respiratory movement during radsurgery,’’ Special Issue on Planning and Image-Guidance in RadiaTherapy, Computer-Aided Surgery, September, 2000.

5S. M. Morrill, M. Langer, and R. G. Lane, ‘‘Real-time couch compenstion for intra-treatment organ motion: Theoretical advantages,’’ MPhys.23, 1083~1996!.

6S. Webb, ‘‘Conformal intensity-modulated radiotherapy~IMRT! deliv-ered by robotic linac-conformality versus efficiency of dose deliverPhys. Med. Biol.45, 1715–1730~2000!.

7S. Webb,The Physics of Three-dimensional Radiation Therapy~Instituteof Physics, Bristol, 1993!.

8H. D. Kubo and B. C. Hill, ‘‘Respiration gated radiotherapy treatmenttechnical study,’’ Phys. Med. Biol.41, 93–91~1996!.

9T. Tadaet al., ‘‘Lung cancer: Intermittent irradiation synchronized witrespiratory motion—results of a pilot study,’’ Radiology207, 779–783~1998!.