In vivo dosimetry by an aSi-based EPIDAngelo Piermattei, Andrea Fidanzio, Gerardina Stimato, Luigi Azario, Luca Grimaldi, Guido D’Onofrio, SavinoCilla, Mario Balducci, Maria Antonietta Gambacorta, Nicola Di Napoli, and Numa Cellini Citation: Medical Physics 33, 4414 (2006); doi: 10.1118/1.2360014 View online: http://dx.doi.org/10.1118/1.2360014 View Table of Contents: http://scitation.aip.org/content/aapm/journal/medphys/33/11?ver=pdfcov Published by the American Association of Physicists in Medicine Articles you may be interested in Portal dose image prediction for in vivo treatment verification completely based on EPID measurements Med. Phys. 36, 946 (2009); 10.1118/1.3070545 Dynamic conformal arc therapy: Transmitted signal in vivo dosimetry Med. Phys. 35, 1830 (2008); 10.1118/1.2900718 Correction of CT artifacts and its influence on Monte Carlo dose calculations Med. Phys. 34, 2119 (2007); 10.1118/1.2736777 Patient dose considerations for routine megavoltage cone-beam CT imaging Med. Phys. 34, 1819 (2007); 10.1118/1.2722470 CT volumetry of the skeletal tissues Med. Phys. 33, 3796 (2006); 10.1118/1.2337272

In vivo dosimetry by an aSi-based EPIDAngelo PiermatteiInstituto di Fisica, Università Cattolica del S. Cuore, Roma, Italy, U.O. di Fisica Sanitaria,Università Cattolica del S. Cuore Policlinico “A. Gemelli,” Roma, Italy and U.O. di Fisica Sanitaria,Centro di Ricerca e Formazione ad Alta Tecnologia nelle Scienze Biomediche dell’Università CattolicaS. Cuore, Campobasso, Italy

Andrea FidanzioU.O. di Fisica Sanitaria, Università Cattolica del S. Cuore Policlinico “A. Gemelli,” Roma, Italy

Gerardina StimatoInstituto di Fisica, Università Cattolica del S. Cuore, Roma, Italy, U.O. di Fisica Sanitaria,Università Cattolica del S. Cuore Policlinico “A. Gemelli,” Roma, Italy and U.O. di Fisica Sanitaria,Centro di Ricerca e Formazione ad Alta Tecnologia nelle Scienze Biomediche dell’Università CattolicaS. Cuore, Campobasso, Italy

Luigi AzarioInstituto di Fisica, Università Cattolica del S. Cuore, Roma, Italy and U.O. di Fisica Sanitaria,Università Cattolica de S. Cuore Policlinico “A. Gemelli,” Roma, Italy

Luca Grimaldi, Guido D’Onofrio, and Savino CillaU.O. di Fisica Sanitaria, Centro di Ricerca e Formazione ad Alta Tecnologia nelle Scienze Biomedichedell’Università Cattolica S. Cuore, Campobasso, Italy

Mario Balducci, Maria Antonietta Gambacorta, Nicola Di Napoli, and Numa CelliniU.O. di Radioterapia, Università Cattolica del S. Cuore Policlinico “A. Gemelli,” Roma, Italy

�Received 13 March 2006; revised 3 July 2006; accepted for publication 12 September 2006;published 31 October 2006�

A method for the in vivo determination of the isocenter dose, Diso, and mid-plane dose, Dm, usingthe transmitted signal St measured by 25 central pixels of an aSi-based EPID is here reported. Themethod has been applied to check the conformal radiotherapy of pelvic tumors and supplies accu-rate in vivo dosimetry avoiding many of the disadvantages associated with the use of two diodedetectors �at the entrance and exit of the patient� as their periodic recalibration and their positioning.Irradiating water-equivalent phantoms of different thicknesses, a set of correlation functions F�w , l�were obtained by the ratio between St and Dm as a function of the phantom thickness, w, for adifferent field width, l. For the in vivo determination of Diso and Dm values, the water-equivalentthickness of the patients �along the beam central axis� was evaluated by means of the treatmentplanning system that uses CT scans calibrated in terms of the electron densities. The Diso and Dm

values experimentally determined were compared with the stated doses Diso,TPS and Dm,TPS, deter-mined by the treatment planning system for ten pelvic treatments. In particular, for each treatmentfour fields were checked in six fractions. In these conditions the agreement between the in vivodosimetry and stated doses at the isocenter point were within 3%. Comparing the 480 dose valuesobtained in this work with those obtained for 30 patients tested with a similar method, which madeuse of a small ion-chamber positioned on the EPIDs to obtain the transmitted signal, a similaragreement was observed. The method here proposed is very practical and can be applied in everytreatment fraction, supplying useful information about eventual patient dose variations due to theincorrect application of the quality assurance program based on the check of patient setup, machinesetting, and calculations. © 2006 American Association of Physicists in Medicine.�DOI: 10.1118/1.2360014�

Key words: in vivo dosimetry, transit dosimetry by EPID

I. INTRODUCTION

Many researchers in the past years have improved methodsto verify the correct dose delivery in patients.1–11 In particu-lar, several groups have examined methods and detectors assemiconductor diodes for in vivo dosimetry. Recently, the useof electronic portal imaging devices �EPIDs� for in vivo do-simetry has also been explored.12–21 However, the unavail-

ability of commercial software for EPIDs does not enabletheir use as stable detectors.22,23

Diodes are the on-line detectors most used today. How-ever, all practical methods for in vivo dosimetry as the onebased on diode measurements of entrance and exit doses re-quire special efforts and many physicists have been discour-aged in verifying the mid-plane dose using these techniques.In vivo entrance dosimetry with diode detectors has been

4414 4414Med. Phys. 33 „11…, November 2006 0094-2405/2006/33„11…/4414/9/$23.00 © 2006 Am. Assoc. Phys. Med.

demonstrated to be a valuable technique to investigate inac-curate x-ray beam delivery.24 However, the entrance dosedoes not supply any information about the dose near thereference point25 in the target volume. Moreover, as is wellknown, there are some problems in the use of diodes such as�i� the necessity of periodic recalibration and correction fortemperature and energy response dependence, �ii� the accu-rate diode positioning and the estimation of the diode photonfluence perturbation, �iii� the time consuming detector posi-tioning at the entrance and exit of the beam central axis forevery gantry angle, and �iv� the two diodes method is inad-equate when asymmetric inhomogeneities along the beamcentral axis are present in the patient.24,26

In a recent paper the authors27 reported a practical methodto estimate the in vivo mid-plane dose, Dm, in patient, usingthe signal St, measured by a small air ion-chamber, posi-tioned on the EPID along the beam central axis, and the CTscan information. The use of a detector positioned below thepatient can avoid the disadvantages related to the diode po-sitioning. Moreover, accurate estimation of the mid-planedose, even in the presence of asymmetric inhomogeneitiesalong the beam central axis, can be obtained considering CTscan information. The proposed technique has been clinicallyapplied in a pilot study for treatments of pelvic tumors byconformed beams. Even if only two of the four fields usedwere tested, due to the small shadow area caused by theion-chamber on the portal vision, the agreement between cal-culated and measured doses was very good.

The availability of a new commercial software �Dosime-try module of the Varian Vision software version 7.3.10 SP3�to read the EPID signals has allowed us to use this device asa portal dosimeter, simplifying the method performance andeliminating the ion-chamber portal imaging perturbations. Inthis work a new algorithm was implemented to supply theisocenter dose in the patient in all the fields of pelvic radio-therapy treatment.

II. MATERIALS AND METHODS

The method here proposed is based on a set of measure-ments carried out by �i� an ion-chamber positioned along thebeam central axis in the geometrical middle point of a ho-mogeneous phantom used to determine the mid-plane dose,Dm, and �ii� 25 central pixels of an aSi-based EPID used tomeasure the transmitted signal, St.

Using different square field size beams and different ho-mogeneous phantom thicknesses, the ratios St /Dm were de-termined for the x-ray beams supplied by a linac Varian Cli-nac 2100 C/D.

A. The aSi-based EPID

In this study we used a commercially available aSi-basedEPID �aS500, Varian Medical System�, mounted on a Clinac2100 C/D that supplied photon beams of 6 and 10 MV. A120 multileaf collimator �MLC� was used to conform theclinical fields.

The EPID system, described in detail elsewhere,28 in-cludes �i� an image detection unit �IDU� featuring the detec-

tor and accessory electronics; �ii� an image acquisition unit�IAS2�, interfacing the hardware that controls and reads theIDU; and �iii� a dedicated workstation �Portal Vision PC�.The IDU is essentially a matrix of 512�384 pixels with aresolution of 0.784�0.784 mm2 and a total sensitive area ofabout 40�30 cm2. Overlying the array is a scintillating layer�gadolinium oxysulphide� and a copper plate �of �1 mmthickness�, making the portal imager an indirect detectionsystem.29,30 The total water-equivalent thickness of the con-struction materials in front of the photodiodes is 8 mm, asspecified by the manufacturer.

A dosimetric module of the Varian Vision software, ver-sion 7.3.10 SP3, enables the EPID calibration in terms ofcalibration units �CU� and the reading of the calibrated signalon the EPID matrix. The EPID calibrated signals are dis-played with an accuracy of 0.001 CU.

Possible saturation effects and their impact on the dosim-etric accuracy are reported in the literature28 as absent ornegligible for all dose rate settings when measurements areperformed at source surface distances �SSDs� equal or largerthan 145 cm. In particular, in this configuration the saturationerror was found to be less than 0.5% for dose rates equal toor less than 400 MU/min.

The linearity of the detector response as a function of thedose and the dose rate is well documented in theliterature.28–31 However, in this work, some measurementswere performed with the EPID at SSD=150 cm and a rangeof MU between 10 and 300 to study the lack of linearity ofthe portal detector signal. The average signal from 25 pixelsaround the center of the EPID �that corresponds to a 3.7�3.7 mm2 area� was examined in terms of day-to-day long-term stability using a 10�10 cm2 field, over a period of6 months. For those measurements the linac output factorvariations were taken into account by using the central ion-chamber of the field detector used for daily dosimetric con-trols. The signal stability was also checked at different gantryangles.

B. Mid-plane dose

The measurements were performed with the 6 and 10 MVx-ray beams, using a homogeneous polymethilmethacrylate�PMMA� phantom made of 25�25 cm2 slabs that allowedus to obtain different phantom thicknesses, z. For the refer-ence configuration, a PTW ion-chamber, model M31010�0.3 cm3 in volume�, positioned at the linac isocenter�SAD=100 cm� and in the center of the PMMA phantom,was used to determine the mid-plane dose, Dm, that is de-fined as the dose to water at the point on the ray-line halfwaybetween the phantom points of entrance and exit of the beamcentral axis �Fig. 1�a��. The water-equivalent depth w /2 ofthe PMMA depth, z /2, at which the ion-chamber was posi-tioned, was obtained multiplying z /2 by the physical PMMAdensities equal to 1.19. The depths w /2 determined with thisprocedure resulted in good agreement within 1.0 mm �thatmeans with a dose discrepancy well within ±1%�, with thedepths obtained applying the method proposed by the TG21.32 For all phantom thicknesses, the measurements were

4415 Piermattei et al.: In vivo dosimetry by an aSi-based EPID 4415

Medical Physics, Vol. 33, No. 11, November 2006

performed with square beams of dimensions l� l, defined atthe SAD equal to 5�5 cm2, 10�10 cm2, 15�15 cm2, and20�20 cm2.

The Dm values were determined applying the practicalcode TRS 398.33 In particular, the ion-chamber reading, MQ,at the quality Q �corrected for influence quantities other thanbeam quality� was multiplied for the calibration factor,ND,w,Q0

, and for the beam quality correction factor, kQ,Q0.

The EPID was positioned at 150 cm from the source, andthe average of the 25 central pixel readings was assumed asthe transmitted signal, St, below the phantom. The presenceof the ion-chamber �used for the Dm measurement� above theEPID central pixels gave a decrease of St that ranged be-tween 0.8% and 1.2% for the different energies and irradia-tion configurations used. Such perturbation was taken intoaccount.

The method here proposed uses empirical correlationfunctions, F�w , l�, obtained by the ratios between the trans-mitted signal, St �CU� �that include the primary and the scat-ter photon component from the EPID�, and the mid-planedose, Dm �Gy�, measured at a geometrical middle point of awater-equivalent phantom of thickness, w �cm�. Using differ-ent square field widths, l �cm�, and five PMMA phantomthicknesses, ranging between 15.0 and 49.0 cm, the ratios

F�w,l� =St

Dm�1�

were determined and fitted with the following expression:

F�w,l� = A�l� + B�l� · C�l�w, �2�

where A, B, and C are fitting parameters dependent on thewidth, l, and w was taken dimensionless. Using the leastsquare method, third-order polynomial functions allowed usto fit the A, B, and C parameters versus l to obtain F�w , l�

values for each field width between 5 and 20 cm.To take into account different irradiation conditions, a set

of measurements �well detailed in a previous paper27� wascarried out positioning the phantom middle point below�mid-down� and above �mid-up� the isocenter.

In particular, letting the isocenter inside the phantom, themiddle point was positioned at distances, d, equal to ±3, ±5,and ±7 cm from the isocenter �d�0 for mid-down and d�0 for mid-up configurations�, and measurements of Dm� andSt�, were obtained using the same phantom thicknesses andfield sizes �defined at the isocenter�, used for the referenceconfiguration �Fig. 1�a��.

For distances, d, between ±7 cm, the percentage differ-ences between the Dm� values corrected for the distance in-verse square law to obtain the dose at the SAD and the Dm

values �measured in reference configuration� showed a quasi-Gaussian-shaped distribution, with a mean shift of 0.1%, andrelative dispersion equal to 0.3% �1 sd�. This result is theo-retically justified in Appendix A.

Moreover, we proposed the following approximation:

St� · f�d,l�

Dm� · �SAD + d

SAD�2 =

St

Dm, �3�

where the f�d ,1� factors take into account the different scat-tered contribution for the St� signal due to the different dis-tances between the EPID and the bottom surface of the phan-tom. These factors were obtained as the ratios between the St

values obtained in the reference configuration and the signalsSt�, obtained for the mid-down or mid-up configurations forevery phantom thickness, w, field width, l, and distance, d.The St /St� ratio for fixed d, and l, resulted independently ofthe phantom thickness �within ±0.3%�. Therefore, for fixedd, and l, the f�d , l� factors were determined as the mean ofthe values obtained for the different phantom thicknessesused.

In conclusion, from Eqs. �1�–�3�, the in vivo mid-planedose Dm can be determined for every configuration by mea-suring the signal, St�, with the equation

Dm =St� · f�l,d�

F�w,l��SAD + d

SAD�2 , �4�

where the F�w , l� and f�d , l� functions have to be relative tothe beam energy and wedge filter used.

Asymmetric inhomogeneous phantoms have been used ina previous paper27 to simulate the presence of bone, and lungtissues. For these phantoms a density middle point was de-fined along the beam central axis as the point of half radio-logical path length �i.e., the point that has over and under thesame phantom thickness in terms of g ·cm−2�. Figure 1�b�shows an example of a phantom with asymmetric inhomoge-neity, where d is the distance between the geometricalmiddle point and the isocenter and d� is the distance betweenthe geometrical middle point and the density middle point.The inhomogeneities were positioned over �Fig. 1�b�� and

FIG. 1. �a� Reference configuration used to irradiate the homogeneousPMMA phantom with the isocenter distance SAD=100 cm �dashed line�coincident with the geometrical middle point ��� of the phantom where theion-chamber �X� was positioned. The geometrical middle point can be be-low �mid-down, +d� or above �mid-up,−d� the isocenter point, respectively.�b� Typical configuration for the simulation of asymmetric inhomogeneities.The shadowed area represents a bone inhomogeneity above the isocenter.The geometrical middle point ��� is at mid-down, +d, from the isocenter,and the position of the ion-chamber �X�, at half thickness, in terms ofg ·cm−2, is at a distance −d� from the geometrical middle point.

4416 Piermattei et al.: In vivo dosimetry by an aSi-based EPID 4416

Medical Physics, Vol. 33, No. 11, November 2006

under the PMMA phantom, obtaining the density middlepoints always inside the PMMA medium. The distance d�was determined by

d� · �̄h =1

2� z

2�̄b −

z

2�̄a ,

d� =z

4�̄h

��̄b − �̄a� , �5�

where �̄a and �̄b are the mean physical density values in thez /2 above the phantom and z /2 below the phantom, respec-tively; �̄h is the mean physical density of the homogeneousmedium �in this simulation the PMMA�. �̄h can be equal tothe �̄a or �̄b values depending on which of the two z /2 mediais homogeneous. The mean physical densities were deter-mined by

�̄ =��izi

z/2, �6�

where �i is the physical density of the ith medium with thick-ness zi.

Then, at the presence of the asymmetric inhomogeneities,Eq. �4� changes into

Dm =St� · f�l,d�

F�w,l��SAD + d + d�

SAD�2 . �7�

The distance, d, in the f�d , l� factor takes into account thereal distance between the bottom phantom surface and theEPID, since f�d , l� was independent of the presence of inho-mogeneities.

In a previous work,27 12 phantoms with asymmetric inho-mogeneity configurations were performed. For photon beamsof 16 and 15 MV, the ratios, R, between the Dm obtained byEq. �7� and the experimental doses Dm� were between 0.98and 1.02 with a standard deviation equal to 0.5%.

C. Isocenter dose

The dose at the isocenter point D�wiso , l� �where wiso is theisocenter depth� in a homogeneous water equivalent phantomwas obtained using the measurements reported in Sec. II Band the tissue maximum ratios �TMRs� of the x-ray beams.34

The TMRs are related with the middle plane dose in phan-tom, D�w /2 , l�, with the dose at depth of maximum dose,D�wmax, l�, and the dose at the isocenter D�wiso , l� by thefollowing expressions:

TMR�w/2,l� =D�w/2,l�D�wmax,l�

, �8�

TMR�wiso,l� =D�wiso,l�D�wmax,l�

. �9�

By the ratios of expressions �8� and �9�

D�wiso,l� = D�w/2,l� ·TMR�wiso,l�TMR�w/2,l�

�10�

is obtained.Rewriting D�w /2 , l�=Dm and D�wiso , l�=Diso and using

Eq. �10� in expression �3�, the following expression is ob-tained:

St

Dm=

St� · f�d,l�

Diso ·TMR�w/2,l�TMR�wiso,l�

. �11�

Using Eq. �1�, Diso is obtained:

Diso =St� · f�d,l�

F�w,l� ·TMR�w/2,l�TMR�wiso,l�

. �12�

D. In vivo dosimetry for conformal beams

The St� in vivo measurements for every beam were used todetermine the Dm and the Diso values. These dose valueswere compared with the middle plane dose computed by theTPS Eclipse version 7.3.10, Dm,TPS, and the stated dose,Diso,TPS.

Figure 2 shows an example of CT scan as reported by theTPS for a patient treated with a PA x-ray beam. The dose,Dm, determined by Eq. �7�, was obtained following threesteps:

�1� the width, l, at the SAD of the equivalent square field ofthe conformed field was computed by the TPS Eclipse;

�2� the CT scan of the patient �generally about 5 mm thick�was used to measure the geometrical patient thicknesson beam central axis, z, and the distance, d, between theisocenter and the geometrical middle point; and

�3� the patient water-equivalent thickness, w, on the beamaxis, was supplied by the TPS Eclipse by means of acalibrated CT scan.35 In the same way the distance, d�,and the water-equivalent depth, wiso, were alsodetermined.

FIG. 2. CT section of a patient treated with a PA x-ray beam and the gantryat 0°. The geometrical middle point at z /2 is at the distance, d, below theisocenter point ��� and at distance, d�, above the density middle point �X�.The planning target volume is reported by a dark line.

4417 Piermattei et al.: In vivo dosimetry by an aSi-based EPID 4417

Medical Physics, Vol. 33, No. 11, November 2006

In conclusion, by using the above mentioned quantities, Eqs.�7� and �12� allowed the determination of the Dm and Diso

values, respectively.In Appendix B the steps that can be followed when the

TPS does not supply directly some of the above quantitiesare reported.

The method capability was tested correcting St� for thelinac output fluctuations �±2% � and for the “belly board”attenuation �1% along the beam central axis� to obtain thesame conditions simulated by the TPS.

The radiotherapy plans were performed by the TPSEclipse, and the dose Dm,TPS and Diso,TPS values, for everybeam, were compared with the Dm and Diso values deter-mined for six treatment fractions. The first three measure-ments were performed sequentially while the followingchecks were performed weekly. Ten pelvic treatments �pros-tate and cervical cancer tumors� were examined testing thebeams with the gantry at 0°, 90°, 180°, and 270°. In particu-lar, 40 fields were tested for six fractions. Therefore, consid-ering two points for every field �the isocenter and the densitymiddle points�, about 480 in vivo dose determinations werecarried out. When the ratios Rm=Dm /Dm,TPS and Riso

=Diso /Diso,TPS were greater than 1.04 or less than 0.96, thephysicist reviewed the patient setup, the machine settings,and the TPS calculations to determine possible errors.

For each field, the mean values, R̄m and R̄iso, and themaximum dispersions of the ratios were determined. More-over, a comparison was carried out between the total mea-sured Diso and computed Diso,TPS for the checked therapyfractions.

III. RESULTS

The mean signal reproducibility of the 25 EPID pixels atthe gantry equal to 0° was estimated to be better than 0.2%�2 sd� and within 0.5% �2 sd� for different gantry angles.Moreover, the long term EPID signal stability �6 months�was within 1%. The linearity with the MUs was within 2%between 30 and 300 MU, while toward 10 MU the signalgradually decreased down to 5%, confirming the literaturedata.28 This nonlinearity was taken into account in the St

determination.Figure 3 shows an example of the experimental ratios,

St /Dm, and the function, F�w , l�, obtained for open and

wedged square beams of different dimensions, l� l, as afunction of the phantom water-equivalent thickness, w, ob-tained for the 10 MV x-ray beam.

Table I reports the coefficients of the third-order polyno-mials that were used to fit the A, B, and C parameters as afunction of the field width, l.

Figure 4 shows the scattering correction factors, f�l ,d��normalized to unity for d=0�, obtained for 6 and 10 MVx-ray beams, field width l=5, 10, 15, 20 cm, and distancesd= ±3, ±5, and ±7 cm. The spread of the factors f�d , l�, forgreater d values was observed to decrease when photon beamenergy increased. In particular, for the negative d values,which are relative to the mid-up condition, St decreases dueto the major distance of the exit phantom surface, while forpositive d values, which are relative to the mid-down condi-tion, St increases due to the minor distance of the exit phan-tom surface. A table of St /St� ratios was used to interpolatethe f�d , l� factors in Eqs. �4�, �7�, and �12�.

The differences between the Dm values obtained by Eq.�4� and the Dm or Dm� values measured in the homogeneousPMMA phantom for the different experimental configura-tions were within ±1%.

Table II summarizes, for the 40 fields examined, theranges of the patient thickness w, the distance d and d�, theequivalent square width, l, and the reference doses at theisocenter Diso,TPS, for single beam and therapy fraction.

TABLE I. Coefficients of the third-order polynomials that were used to fit the A, B, and C parameters for the 6 and 10 MV photon beams. In particular, thethird-order polynomials were written in the form A�l�=a0+a1x+a2x2+a3x3, B�l�=b0+b1x+b2x2+b3x3, and C�l�=c0+c1x+c2x2+c3x3.

Beam 6 MV 10 MV

a0 a1 a2 a3 a0 a1 a2 a3

A 7.06�10−2 −4.54�10−3 6.05�10−4 −1.43�10−5 7.15�10−2 7.97�10−3 −3.45�10−4 6.21�10−6

b0 b1 b2 b3 b0 b1 b2 b3

B 2.99�10−1 7.16�10−5 1.00�10−4 −3.37�10−6 2.97�10−1 −7.08�10−3 6.84�10−4 −1.67�10−5

c0 c1 c2 c3 c0 c1 c2 c3

C 9.68�10−1 1.90�10−4 −1.00�10−4 3.15�10−6 9.79�10−1 −2.93�10−3 1.40�10−4 −2.99�10−3

FIG. 3. Experimental St /Dm ratios and the fits �continuous lines� obtainedfor the 10 MV open and wedged x-ray beams of different field dimensions,5�5 ���, 10�10 ���, 15�15 ���, and 20�20 cm2 ���, as a function ofthe phantom water-equivalent thickness, w.

4418 Piermattei et al.: In vivo dosimetry by an aSi-based EPID 4418

Medical Physics, Vol. 33, No. 11, November 2006

For the AP or PA irradiation, the patients showed bonetissue asymmetrically positioned, while for the LL irradia-tions, the bone tissue was symmetrically positioned. The Diso

and Dm values, determined by the in vivo procedure, werenormalized to the Diso,TPS and, Dm,TPS, respectively, to obtainthe ratios Riso and Rm. The Dm,TPS values generally equal toDiso within 5%, and the distances d+d� were generally lessthan 2 cm. Using a computer spreadsheet program, the de-termination of these ratios required a minimal workload.

Ratios Rm or Riso ranging between 0.93 and 0.96 wereobserved for two patients when treated with PA irradiations.This result was due to the presence of gas cavities in thepatient’s abdomen as recorded in the CT scan. The cavitydimensions ranged between 3 and 5 cm along the beam cen-tral axis and were positioned between the source and theisocenter point. New CT scans were obtained without the gaspockets, and the ratios Riso that resulted were well within therange between 0.96 and 1.04. Figure 5 shows the Riso valuesobtained for one of these two patients. The first two fractions

show, for the 0° and 180° gantry, a large difference betweenthe Diso and Diso,TPS values. The successive six fractionsshow a good agreement between the Diso and Diso,TPS valuesobtained by a new CT scan without gas cavities.

Figure 6 shows the histogram of the Riso and Rm valuesobtained for the 40 fields examined where the dimensions ofthe gas cavities were less than 2 cm.

FIG. 4. f�d , l� factors for 5�5 cm2 ���, 10�10 cm2 ���, 15�15 cm2 ���,and 20�20 cm2 ��� square fields of 6 MV �a� and 10 MV �b� x-ray beamsas a function of the distance, d. The negative d values are relative to themid-up condition while the positive d values are relative to the mid-downcondition.

TABLE II. Ranges of some parameters observed for the ten patients exam-ined. The thickness w, the distances d and d�, the equivalent square field l,and the stated dose Diso,TPS for beam fractions are reported.

w �cm� d �cm� d� �cm� l �cm� Diso,TPS �Gy�

18–39 ±2 ±1 5–16 0.3–0.7

FIG. 5. Ratios Riso obtained for eight fractions of a pelvic tumor treated withfour beams at gantry 0° ���, 180° ���, 90° ���, and 270° ���. For the firsttwo fractions we used the computed Diso,TPS by a CT scan that presented alarge gas cavity along the beams at 0° and 180°. For the successive fractionsthe Diso,TPS was obtained by a CT scan without gas cavities.

FIG. 6. The 480 ratios Riso �white� and Rm �gray� obtained in this work aretogether with the 240 Rm ratio values �black�, obtained by a previous work.27

The mean value of the ratios is equal to 1.00±0.05 �2 sd�.

4419 Piermattei et al.: In vivo dosimetry by an aSi-based EPID 4419

Medical Physics, Vol. 33, No. 11, November 2006

For every beam, the maximum dispersions, �, of theseratios �between 4% and 6%� were symmetrical enough withrespect to the mean values. The ratios are reported with the240 Rm values obtained in the previous work where a smallion-chamber on the EPID27 was used. In that work, the meanvalue of the Rm values was equal to 1.01±0.05 �2 sd�, whilethe 480 Riso and Rm values obtained in this paper show amean value equal to 0.99±0.05 �2 sd�. The mean value of720 ratios is equal to 1.00±0.05 �2 sd�.

Figure 7 shows the ratios between the total doses at theisocenter point, measured by the in vivo method and com-puted by the TPS during the total number of the therapyfractions tested. In particular, the results obtained for the tenpatients examined were compared with the 30 sets of dataobtained in the previous paper.27 In that work we reportedonly the dosimetry at the middle point. The implementationof the procedure here reported allowed us to determine theDiso also for those patients. It is important to underline thatthose 30 patients �that did not present gas cavities greaterthan 2 cm� were tested with only two of the four fields �0°and 90°� and the small number of fields examined could bethe cause of major dispersion of the ratios reported in Fig. 7.For the two patients that presented large gas cavities, theresults that could have been obtained without a new CT scanare reported. The presence of these cavities with dimensionsin the range of 3–5 cm supply a discrepancy between Diso

and Diso,TPS up to 7% for PA irradiations, while the totaldoses at the isocenter point obtained by the four fields re-sulted in disagreement up to 4%.

IV. DISCUSSIONS

As reported in the literature,36 the ultimate check of theactual dose delivered to a patient can only be performed atthe patient level, by means of in vivo dosimetry. When aQ.A. program is followed, the percentage of patients witherrors larger than 5% in dose delivery can be reduced to 1%.If not, this number will increase to 3%–10% or even higher

values. In conclusion, practical in vivo tests are needed toverify that the Q.A. programs are correctly applied.

In a previous work27 the authors have proposed a practicalmethod for in vivo dosimetry that uses a small ion-chamberpositioned on the EPID. The same method is here imple-mented �i�, using the signal of the 25 central pixels of anaSi-based EPID for the determination of both the mid-planedose and the dose at the isocenter point. The use of the EPIDreduces again the efforts required by detector positioning.Moreover, the control of the dose at the isocenter point doesnot require new TPS calculations when the point coincideswith the reference point of the treatment where the dose forevery single beam is stated.

The measurements required by the method are �i� a set ofcorrelation functions F�w , l� for open and wedged x-raybeams measured with a water-equivalent phantom, position-ing its mid-plane at the SAD, and �ii� a set of f�d , l� factorsthat take into account the variation of the x-ray scatter con-tribution for St� due to the different distances between theEPID and the phantom bottom surface. These last factorsslowly change with d and l and were independent of thephantom thickness.

The efficacy of the inverse square law in the range of thed values examined, as well as the small variations of thephoton scatter component from the phantom to the EPID,allow an accuracy of 1% for the Dm determination by Eq. �4�in a homogeneous phantom positioned in the different con-figurations �iso-up and iso-down�. This accuracy also in-cludes the approximation of the fits used.

Ten pelvic treatments, with the bone tissue asymmetri-cally distributed for AP or PA irradiations, and bone tissuesquasi-symmetrically distributed for LL irradiations, were ex-amined. The patient’s thicknesses were determined using CTscans calibrated in terms of electron densities. For everyfield, the two dose-points were measured and the ratios withthe doses calculated by the TPS were determined.

The Dm,TPS values resulted in many cases within 5% ofthe Diso,TPS that was chosen to coincide with the dose at theICRU reference point.25 The ratios Riso and Rm reported inFig. 6 showed a very similar distribution, which means thatthe accuracies of Eqs. �7� and �12� are of the same order. Theagreement between planned and measured dose values werewithin ±5% for more than 95% of the examined fields. Thisagreement includes the uncertainty of the method �2%� in aninhomogeneous phantom, the uncertainties in the determina-tion of patient’s thicknesses �including the presence of smallgas cavities�, and other uncertainties such as the equivalentsquare field determination and the St reproducibility.

When the disagreement between planned and measureddose values was higher than 4%, the physicist reviewed thepatient setup, the machine settings, and the TPS calculationsto detect possible errors.

The presence of gas pockets in the abdomen located alongthe beam central axis of the AP or PA irradiation was ob-served in the CT scans of some patients. Two of them pre-sented the diameters of the cavities greater than 3 cm, andthe ratio, R, obtained for PA or AP irradiations was out of the

FIG. 7. Ratios between the total doses at the isocenter point, measured bythe in vivo method and computed by the TPS. The ten ratios obtained in thiswork ��� with the four fields examined are reported together with 30 ratiosobtained in a previous work27 with only two fields ���. Two ratios ��� oftwo patients that presented in the CT scan large gas cavities are reported,too.

4420 Piermattei et al.: In vivo dosimetry by an aSi-based EPID 4420

Medical Physics, Vol. 33, No. 11, November 2006

tolerance level. This result can be explained considering thatthe patient’s water equivalent thickness, w, obtained by theCT scan in the presence of a gas pocket, is lower than theone estimated for the same patient without gas cavities. Ob-serving Fig. 3, the function F�w , l� increases for lower wthicknesses. Therefore, in successive therapy fractions, whenthe gas cavity is absent, the Dm is determined by Eq. �4� or�7� with an overestimated F�w , l� value and a lower value ofSt �this last value is due to the additional tissue present alongthe beam central axis with respect to the case with a gaspocket�. Therefore, when the gas pockets are absent, the invivo Dm value is lower than the Dm,TPS value determined fora CT scan with gas cavity. These results suggested to reducethe presence of the gas cavities in the patient or to acceptcavities with dimensions less than 2 cm.

The results reported in Fig. 7 show that the method is ableto verify an agreement within ±3% between the stated and invivo total measured doses at the isocenter point, when anadequate Q.A. program is applied. Such good results canalso be obtained using a method based on entrance and exitdiode positioning. However the frequent diode calibrations,as well as the diode positioning on the patient �where portalimages are recommended to check the diode position37�, re-quire more workload.

V. CONCLUSIONS

In vivo dosimetry is currently still applied in a relativelysmall number of centers. Many researchers are today study-ing new methods and new adequate on-line detectors toverify with major accuracy the correct dose delivery into thepatient. The method here reported was previously applied27

to determine the mid-plane dose of a group of patientstreated for pelvic tumors. In that work the transmitted signalSt was obtained by a small ion-chamber positioned on theEPIDs. In this work St was obtained by the central pixels ofan aSi-based EPID, calibrated by a commercial software�Dosimetry module of the Varian Vision software version7.3.10 SP3�. The use of EPIDs strongly reduced the work-load of the in vivo dosimetry. Moreover, the method is nowable to supply the dose at the isocenter point.

The measurements needed to implement Eqs. �7� and �12�are not excessive so the data required to apply the method toa single photon beam can be acquired with 140 measure-ments of St and Dm in about 4 h. It is also possible to calcu-late the Dm values by a TPS instead of measuring them.However, this solution simplifies the measurements but doesnot significatively reduce the time because St and Dm aremeasured at the same time.

The Dm and Diso computation has been carried out on acomputer spreadsheet program to minimize the workload.

The method seems to be a good ancillary program to takeinto account the dosimetric variations that depended on thesmall exceeding of the Q.A. tolerance level. Indeed, the re-sults of this and of a previous paper27 underline the impor-tance of the patient setup verification �due to the changes ofpatient morphology during the treatment� and confirm thatthe method here proposed is a practical test to verify the

consistency of Q.A. programs. We are currently studying theeffect of the gas cavity dimensions also outside the beamcentral axis. As pointed out in the literature,37 the gas cavitiesare the causes of major dispersion of the ratios between mea-sured and prescribed doses for pelvic tumor treatments.

APPENDIX A

In Fig. 1, Dm and Dm� measured in a phantom at depth w /2using a square field width, l and l�, respectively, can be writ-ten by

Dm�w/2,l� = Dm,0TAR�w/2,l� , �A1�

Dm� �w/2,l�� = Dm,0� TAR�w/2,l�� , �A2�

where Dm,0 and Dm,0� are the doses in a water-equivalent el-ement �that is sufficient for the charge particle equilibrium�obtained in free space at distances from the source equal to100 cm and 100+d cm, and TARs are the tissue air ratios.

Determining the TAR as a product of the tissue maximumratio �TMR� and the peak scatter factor �PSF�34 for d valueswithin ±7 cm, the agreements between TAR�w /2 , l� andTAR�w /2 , l�� were well within 0.7%.

Introducing the approximation

Dm,0 Dm,0� �SAD + d

SAD�2

, �A3�

the ratio between Eqs. �A1� and �A2� supplies

Dm Dm� �SAD + d

SAD�2

. �A4�

APPENDIX B

The parameters reported in Sec. II D for the determinationof the Dm and Diso can be obtained following these steps.

�1� For conformal pelvic fields the width, l, of the equiva-lent square field can be obtained by the Sterlingapproximation.38,39,27

�2� The CT scan of the patient �generally about 5 mmthick�, which contains the beam central axis, can be usedto measure, on this axis the geometrical patient thick-ness, z, and the distance, d, between the isocenter andthe geometrical middle point at thickness z /2.

�3� Two rectangular strips can be drawn on the CT scan ofthe patient �Fig. 2�, one over the other, z /2 high andabout 5 mm wide, centered on the beam central axis todetermine the mean densities, �̄a and �̄b.

�4� Another rectangular strip can be drawn on the CT scan,centered on the beam central axis, drawn up to the depthziso, to determine the mean density, �̄iso.

�5� The TPS calibrated CT numbers can be used to deter-mine the mean electronic density of the two rectangularstrips. Next, the mean physical densities, �̄a and �̄b, areestimated by the linear relation between the electronicdensity and the physical density of human tissues, ob-tained by the stoichiometric calibration of the Houn-sfield numbers.35 These last values resulted very little

4421 Piermattei et al.: In vivo dosimetry by an aSi-based EPID 4421

Medical Physics, Vol. 33, No. 11, November 2006

dependence on the width of the rectangular strips thatchanged between 3 and 7 mm and were in good agree-ment with the values obtained directly by the EclipseTPS.

�6� The water-equivalent thickness of the patient was deter-mined by

w =z��̄a + �̄b�

2�water�B1�

and the water-equivalent depth of the isocenter pointwas determined by

wiso = ziso�̄iso

�water. �B2�

�7� For an asymmetric inhomogeneity the mean density val-ues, �̄a and �̄b, and �̄h were used to determine the dis-tance, d�, by Eq. �5�. Observing Fig. 2, �̄h in patient is avalue near to one, due to the presence of soft tissue inthe rectangular strip.

�8� For quasi-symmetric inhomogeneities �LL irradiation inFig. 2� d� can be determined by Eq. �5� with �̄h equal toone, due to the presence of soft tissue near the patient’sgeometrical middle point. In these cases d� is a fewmillimeters.

1R. Boellaard, M. van Herk, and B. J. Mijnheer, “A convolution model toconvert transmission dose images to exit dose distributions,” Med. Phys.24, 189–199 �1997�.

2S. Broggi, C. Fiorino, and R. Calandrino, “A simple and robust methodfor in vivo midline dose map estimations using diodes and portal detec-tors,” Radiother. Oncol. 58, 169–178 �2001�.

3M. Essers et al., “The accuracy of CT-based inhomogeneity correctionsand in vivo dosimetry for the treatment of lung cancer,” Radiother. Oncol.37, 199–208 �1995�.

4M. Essers, J. H. Lanson, and B. J. Mijnheer, “In vivo dosimetry duringconformal therapy of prostatic cancer,” Radiother. Oncol. 29, 271–279�1993�.

5V. N. Hansen, P. M. Evans, and W. Swindell, “The application of transitdosimetry to precision radiotherapy,” Med. Phys. 23, 713–721 �1996�.

6S. Heukelom, J. H. Lanson, and B. J. Mijnheer, “In vivo dosimetry duringpelvic treatment,” Radiother. Oncol. 25, 111–120 �1992�.

7S. Heukelom, J. H. Lanson, and B. J. Mijnheer, “Quality assurance of thesimultaneous boost technique for prostate cancer: Dosimetric aspects,”Radiother. Oncol. 30, 66–73 �1994�.

8P. D. Higgins, P. Alaei, B. J. Gerbi, and K. E. Dusenbery, “In vivo diodedosimetry for routine quality assurance in IMRT,” Med. Phys. 30, 3118–3123 �2003�.

9P. A. Jursinic, “Implementation of an in vivo diode dosimetry programand changes in diode characteristics over a 4-year clinical history,” Med.Phys. 38, 1718–1726 �2001�.

10G. Leunens, J. Verstraete, and A. Dutreix, “Assessment of dose inhomo-geneity at target level by in vivo dosimetry: can the recommended 5%accuracy in the dose delivered to the target volume be fulfilled in dailypractice?” Radiother. Oncol. 25, 242–250 �1992�.

11A. Noel, P. Aletti, P. Bey, and L. Malissard, “Detection of errors in indi-vidual patients in radiotherapy by systematic in vivo dosimetry,” Radio-ther. Oncol. 34, 144–151 �1995�.

12R. Boellaard, M. van Herk, and B. J. Mijnheer, “The dose response rela-tionship of a liquid-filled electronic portal imaging device,” Med. Phys.23, 1601–1611 �1996�.

13M. Essers, R. Boellaard, M. van Herk, J. H. Lanson, and B. J. Mijnheer,“Transmission dosimetry with a liquid-filled electronic portal imagingdevice,” Int. J. Radiat. Oncol., Biol., Phys. 34, 931–941 �1996�.

14M. Essers, B. R. Hoogervorst, M. van Herk, J. H. Lanson, and B. J.Mijnheer, “Dosimetric characteristics of a liquid-filled electronic portalimaging device.” Int. J. Radiat. Oncol., Biol., Phys. 33, 1265–1272

�1995�.15D. Huyskens, J. Van Dam, and A. Dutriex, “Midplane dose determination

using in vivo dose measurements in combination with portal imaging,”Phys. Med. Biol. 39, 1089–1101 �1994�.

16R. J. W. Louwe, E. M. F. Damen, M. van Herk, A. W. H. Minken, O.Torzsok, and B. J. Mijnheer, “Three-dimensional dose reconstruction ofbreast cancer treatment using portal imaging,” Med. Phys. 30, 2376–2389 �2003�.

17P. N. McDermott, “The physical basis for empirical rules used to deter-mine equivalent fields for phantom scatter,” Med. Phys. 25, 2215–2219�1998�.

18L. N. McDermott, M. Wendling, J. Sonke, B. van Asselen, M. van Herk,and B. Mijnheer, “First clinical experience with pre-treatment and in vivoIMRT verification using EPID dosimetry,” Med. Phys. 32, 2090 �2005�.

19T. R. McNutt, T. R. Mackie, P. Reckwerdt, and B. R. Paliwal, “Modelingdose distribution from portal dose images using the convolution/superposition method,” Med. Phys. 23, 1381–1392 �1996�.

20M. Partridge, M. Ebert, and B. M. Hesse, “IMRT verification by three-dimensional dose reconstruction from portal beam measurements,” Med.Phys. 29, 1847–1858 �2002�.

21W. J. C. van Elmpt, M. J. J. G. Nijsten, B. J. Mijnheer, and A. W. H.Minken, “Experimental verification of a portal dose prediction model,”Med. Phys. 32, 2805–2818 �2005�.

22K. L. Pasma, M. Kroonwijk, A. G. Visser, and B. J. M. Heijmen, “Portaldose measurements with a video-based electronic portal imaging deviceusing a deconvolution algorithm,” in Proceedings of the XIIth Interna-tional Conference on the Use of Computers in Radiotherapy, Salt LakeCity, UT �1997�, pp. 282–284.

23A. Piermattei et al., “Verification of computed portal doses by a lineararray of liquid ion-chambers,” Phys. Medica XX�3�, 111–119 �2004�.

24ESTRO, “Practical guidelines for the implementation of in vivo dosime-try radiotherapy with photon beams �entrance dose�,” Physics for ClinicalRadiotherapy, Booklet No. 5 �2001�.

25ICRU, “ICRU Report 50: Prescribing, recording, and reporting photonbeam therapy” �Bethesda, Maryland, International Commission on Radia-tion Units and Measurements, 1993�.

26M. Essers and B. J. Mijnheer, “In vivo dosimetry during external photonbeam radiotherapy,” Int. J. Radiat. Oncol., Biol., Phys. 43, 245–259�1999�.

27A. Piermattei et al., “In-vivo portal dosimetry by an ionization chamber,”Physica Medica XXI, 143–151 �2005�.

28A. Van Esch, T. Depuydt, and D. P. Huyskens, “The use of an aSi-basedEPID for routine absolute dosimetric pre-treatment verification of dy-namic IMRT fields,” Radiother. Oncol. 71, 223–234 �2004�.

29Y. El-Mohri et al., “Relative dosimetry using active matrix flat-panelimager �AMFPI� technology,” Med. Phys. 26, 1530–1541 �1999�.

30P. Munro and D. C. Bouius, “X-ray quantum limited portal imaging usingamorphous silicon flat-panel arrays,” Med. Phys. 25, 689–702 �1998�.

31B. M. C. McCurdy, K. Luchka, and S. Pistorius, “Dosimetric investiga-tion and portal dose image prediction using an amorphous silicon elec-tronic portal imaging device,” Med. Phys. 28, 911–924 �2001�.

32AAPM TG 21, “A protocol for determination of absorbed dose for highenergy photon and electron beams,” Med. Phys. 10, 741–771 �1983�.

33IAEA �International Atomic Energy Agency�, “Absorbed dose in externalbeam radiotherapy: an international code of practice for dosimetry basedon standards of absorbed dose to water,” IAEA TRS 398 �2001�.

34BJR supplement 25, “Central axis depth dose data for use in radio-therapy,” �British Institute of Radiology, London, 1996�.

35U. Schneider, E. Pedroni, and A. Lomax, “The calibration of CT Houn-sfield units for radiotherapy treatment planning,” Phys. Med. Biol. 41,111–124 �1996�.

36C. Weltens, J. Van Dam, and G. Leunens, “Reliability of clinical port formeasuring dose inhomogeneities in radiotherapy for head and neck tu-mours,” Radiother. Oncol. 30, 167–170 �1994�.

37J. H. Lanson, M. Essers, G. J. Meijer, A. W. H. Minken, G. J. Uiterwaal,and B. J. Mijnheer, “In vivo dosimetry during conformal radiotherapyRequirements for and findings of a routine procedure,” Radiother. Oncol.52, 51–59 �1999�.

38T. D. Sterling, H. Perry, and L. Katz, “Automation of radiation treatmentplanning,” Br. J. Radiol. 37, 544–550 �1964�.

39B. E. Bjärngard and R. L. Siddon, “A note on equivalent circles, squares,and rectangles,” Med. Phys. 9, 258–260 �1982�.

4422 Piermattei et al.: In vivo dosimetry by an aSi-based EPID 4422

Medical Physics, Vol. 33, No. 11, November 2006