An Electromagnetic Head and Neck Hyperthermia Applicator: experimental phantom verification and FDTD model

Int. J. Radiation Oncology Biol. Phys., Vol. 68, No. 2, pp. 612– 620, 2007 Copyright © 2007 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/07/$–see front matter

doi:10.1016/j.ijrobp.2007.01.035

PHYSICS CONTRIBUTION

ELECTROMAGNETIC HEAD-AND-NECK HYPERTHERMIA APPLICATOR: EXPERIMENTAL PHANTOM VERIFICATION AND FDTD MODEL MARGARETHUS M. PAULIDES, M.SC., JURRIAAN F. BAKKER, B.SC.,

AND

GERARD C.

VAN

RHOON, PH.D.

Department of Radiation Oncology, Section of Hyperthermia, Daniel den Hoed Cancer Center, Erasmus Medical Center, Rotterdam, The Netherlands Purpose: To experimentally verify the feasibility of focused heating in the neck region by an array of two rings of six electromagnetic antennas. We also measured the dynamic specific absorption rate (SAR) steering possibilities of this setup and compared these SAR patterns to simulations. Methods and Materials: Using a specially constructed laboratory prototype head-and-neck applicator, including a neck-mimicking cylindrical muscle phantom, we performed SAR measurements by electric field, Schottkydiode sheet measurements and, using the power-pulse technique, by fiberoptic thermometry and infrared thermography. Using phase steering, we also steered the SAR distribution in radial and axial directions. All measured distributions were compared with the predictions by a finite-difference time-domain– based electromagnetic simulator. Results: A central 50% iso-SAR focus of 35 ⴞ 3 mm in diameter and about 100 ⴞ 15 mm in length was obtained for all investigated settings. Furthermore, this SAR focus could be steered toward the desired location in the radial and axial directions with an accuracy of ⬃5 mm. The SAR distributions as measured by all three experimental methods were well predicted by the simulations. Conclusion: The results of our study have shown that focused heating in the neck is feasible and that this focus can be effectively steered in the radial and axial directions. For quality assurance measurements, we believe that the Schottky-diode sheet provides the best compromise among effort, speed, and accuracy, although a more specific and improved design is warranted. © 2007 Elsevier Inc. Hyperthermia, Dynamic-specific absorption rate, SAR, Measurements, Antenna array, Head-and-neck tumors.

Locoregional control of advanced carcinoma in the headand-neck region still poses a major therapeutic challenge (1, 2), and treatment-related toxicity remains a major issue in this region (3). Combining hyperthermia (HT) with the current treatment modalities has a high potential to increase positive treatment outcomes without increasing toxicity (4). Recent Phase III trials have shown the benefit of the addition of HT to current treatment modalities for various tumor sites (5–7). Valdagni et al. (8, 9) have demonstrated that with a nonsitespecific HT applicator a significant increase in local control (from 24% for radiotherapy alone to 69% for radiotherapy plus HT) can be achieved for metastatic lymph nodes in Stage IV head-and-neck carcinoma. Extension of the HT application to deeper located tumors is cumbersome with the available applicators; therefore, a specifically designed

applicator is needed that is also expected to contribute to a better treatment quality for this site (10, 11). In recent years, we have investigated the ability of a multi-element circumferential head-and-neck HT applicator to deposit radiofrequency energy selectively to the center of the neck (12). The optimal frequencies were shown to be in the frequency band between 300 and 600 MHz for all realistic target sizes; therefore, we selected 433 MHz as the operating frequency (433 MHz is an ISM frequency [i.e., a frequency available for industry, science, and medicine]). Furthermore, we found that eight antenna elements provided sufficient focusing for selective heating. Building on these findings, we analyzed the influences of the anatomy on the dynamic specific absorption rate (SAR) pattern (13). In this study, we found that muscle phantom measurements are predictive for the SAR pattern in the middle of the neck.

Reprint requests to: Margarethus M. Paulides, M.Sc., Department of Radiation Oncology, Section of Hyperthermia, Daniel den Hoed Cancer Center, Erasmus Medical Center, P.O. Box 5201, Rotterdam NL-3008 AE The Netherlands. Tel: (⫹31) 10-4391676; Fax: (⫹31) 10-439-1022; E-mail: [email protected] Supported by the Maurits and Anna de Kock Foundation and the Dutch Cancer Society, Grant DDHK 2003-2855. Conflict of interest: none, other than a concurrent patent application.

Acknowledgments—We would like to thank the workshop of the Department of Radiation Oncology for the construction of the prototype and T.A. ter Meer, H. Schippers, and W. Schaap of the Dutch Aerospace Laboratory (NLR) for their work on the signal measurement equipment. Also, we thank H. Strijbos and A.P.M. Zwamborn of TNO (The Hague) for their contributions to the discussions. Received Oct 12, 2006, and in revised form Jan 8, 2007. Accepted for publication Jan 8, 2007.

INTRODUCTION

612

Head-and-neck applicator prototype: verification and model

Because of the spine, the central SAR focus changes from a circular to a more donut shape, but the central focus remains at its central position. Subsequently, research was conducted into the possibilities of further improving the power absorption pattern by going from one to two or three antenna rings (14), which also provided axial steering possibilities. We established that heating at depth correlated highly with the amount of antennas arranged in multiple rings. Multiple rings enabled steering in the direction of the patient axis, which has been well validated for deep HT (15, 16) but needs confirmation for the head-and-neck application. This work resulted in a final setup design of a setup with two rings of six antennas, each operating at 433 MHz. For radiators, we selected a patch antenna setup that is simple, low cost, efficient, and small, and we designed it specifically for the head-and-neck applicator (17). Another major advantage of this antenna design was that no matching network is required, which facilitates better control of the input parameters of the setup and hence provides better possibilities to create a representative model. Subsequently, the latter improved the reliability and accuracy of the numeric predictions. The aim of the present study was to experimentally verify the feasibility of focused heating in the head-and-neck region using an array of two rings of six antennas by the split-phantom technique. Simultaneously, we established the accuracy of the predictions of our electromagnetic (EM) simulation model. Finally, we explored the dynamic SAR steering behavior of the antenna array. Then, we constructed a laboratory prototype applicator consisting of 12 patch antenna elements and a muscle-equivalent neck phantom. We performed measurements with this setup and compared them with the predicted SAR patterns. METHODS AND MATERIALS Prototype design Figure 1 shows a picture of the bottom half of the prototype applicator, and Fig. 2 gives an exploded view of its two parts. These parts can be separated to allow two-dimensional infrared (IR) measurements or positioning of measurement devices: glass

y

●

M. M. PAULIDES et al.

z

x

613

Water outflow Water inflow Array 2 Array 1 Phantom

E-field direction Water outflow

Array 2 Array 1

Water inflow Fig. 2. Exploded view of the prototype constructed in two parts for measurements (temperature probes, electric field sheet, infrared pictures). Visible are phantom semicylinders (pink/light gray), two frames spanning sheet, polyvinyl chloride backplanes (dark gray), and copper tape on inside of backplanes (light gray).

fiber probes or electric field Schottky-diode sheet. Both parts of the applicator consisted of a semicircular polyvinyl chloride (PVC) backplane (diameter, 30 cm; length, 60 cm; Fig. 3) and PVC shields at both ends. Both half backplanes were covered and waterproofed, with a 50-␮m-thick ABS copolymer sheet glued to a frame that can be mounted on the backplanes (Fig. 2). Twelve antenna elements were mounted on the PVC backplanes in a 2 ⫻ 6 arrangement (14), that is, two rings (distance, 8 cm; ring radius, 30 cm; Fig. 3) of six equidistantly spaced patch antennas. We defined the origin of the applicator between the rings so the antennas in these rings were located at z ⫽ ⫺4 cm and z ⫽ 4 cm. The water in the prototype was circulated and temperature controlled by two E4850 refrigerated recirculators (Bio-Rad, Microscience Division, Cambridge, MA). Every patch antenna (17) consisted of a resonant, 2-mm-thick, brass patch (width, 12 mm; length, 25 mm) that was asymmetrically (the distance from the rod to the side of the patch was 0.5 mm) mounted on the extending conducting rod of a C-female receptacle connector. Coaxial cables were used to transfer the 433.92-MHz signal from the PG70.150.2 power generators (SSB Electronic, Germany) to the applicator. The conducting backplane required for the patch antenna design was formed by soft copper tape (thickness, 35 ␮m plus 25-␮m adhesive) that covered the inner side of both PVC backplanes (Fig. 1). Two PVC semicylinders (muscle cylinder diameter, 11.6 cm, PVC thickness, 1.5 mm) filled with muscle-equivalent material, representing the neck of a patient, were positioned at the center of the prototype. (The dimensions of the phantom were approximately equal to an average neck [13]). The muscle-equivalent material was created using a recipe described by Ito et al. (18) and had properties in accordance with Gabriel et al. (19) (Table 1).

EM modeling

Fig. 1. Bottom half of prototype with sheet and its frame excluded. Visible are copper tape that covers backplane, one-half of muscle phantom, and multiple patch antennas.

We predicted the SAR distributions in the prototype applicator using the simulation platform for EMC, antenna design, and dosimetry (SEMCAD) (20). SEMCAD has a finite-difference timedomain (FDTD) kernel that can be used for dosimetric simulations. It includes a solid modeling kernel in which we created a full three-dimensional model using the exact dimensions of the applicator prototype and muscle phantom. The measurements of the very thin sheets between the semicylinders were neglected in the calculations. Furthermore, instead of modeling the complete connector and the transmission line, we used the commonly accepted

614

I. J. Radiation Oncology

●

Biology

●

Physics

Volume 68, Number 2, 2007

Fig. 3. Cross-section through prototype at y ⫽ 0 cm with dynamic-specific absorption rate distribution obtained from two infrared pictures (before and after heating) for central settings (a) and zoomed version of this dynamic-specific absorption rate distribution (b). In Fig. a, location of Schottky diodes (electric field sheet 1 and 2) indicated by black rectangles and extent of sheets by dotted lines. Location of antenna rings and main dimensions of prototype and phantom indicated. Temperature probe measurements taken at locations on “x ⫽ 0” and “z ⫽ 0”. In Fig. b, color scale is from black (0%) to white (100%), and length (L50%) and width (W50%) of focus (i.e., 50% iso-dynamic-specific absorption rate contour, indicated.

approach of applying a voltage source at the gap between the conducting backplane and the rod of the patch antenna (21), which provides accurate field distributions (22). This solid model was converted automatically into a grid implementation with maximal grid steps of ␭/15 and grid refinements at the locations of the antennas (␭/250 ⬍ ⌬ ⬍ ␭/80; i.e., a total of ⬃5 M cells). The grid refinements at the antennas were required to accurately model the (important) small dimensions of the antenna and the nonorthogonally placement in the grid, at the location of the feeding rod. Table 1 gives the properties of all the materials used in the simulation. The properties of de-mineralized water (salinity, 0.04 g/L; temperature, 25°C) were found in a publication of Stogryn (23). We exited the voltage sources at all gaps at 433 MHz, and a steady state was obtained after ⬃25 periods.

Measurements Three SAR pattern measurement methods were used, because they all had their own features. Their major strengths were either accuracy (Fiberoptic thermometry), resolution (IR thermography), or speed (electric field, Schottky-diode sheet). These measurement methods and their results were compared to establish the optimal method for quality assurance. Water bolus temperatures were chosen such that the mean temperatures in the prototype were about 22°C, at which the phantom properties are reported i.e., the water bolus temperature was 18°–19°C for the power pulse measurements (temperature probes and IR) and 22°C for the electric field measurements. Figure 3 shows a cross-section through the

Table 1. Properties of the materials at 433 MHz Material

␳ (kg/m3)

␧r (⫺)

␴eff (S/m)

Muscle phantom PVC Demineralized water

900 1,350 1,000

56 2.2 78

0.8 0.004 0.04

Abbreviations: ␳ ⫽ mass density; ␴eff ⫽ effective conductivity; PVC ⫽ polyvinyl chloride; ␧r ⫽ relative permittivity.

applicator at y ⫽ 0 cm and a zoomed version of the SAR distribution obtained by taking IR pictures before and after heating. Fiberoptic thermometry. The temperatures were measured with 24 Takaoka FTP1 standard sensor probes (accuracy, 0.1°C; precision, 0.1°C), which were read with a Takaoka FT1310 fiber thermometer (Takaoka Electric MFG, Japan) with a refresh rate of 3 s. These nonmetallic probes were positioned at the main axes between both halves of the prototype. The local SAR was measured using the power pulse method—the change in temperature (3–10°C) in a phantom was measured while applying a high-power (300 – 600 W) for a short time (90 –300 s) at the antennas of the applicator (24) (Fig. 4). To reduce influences of heat conductivity, a short heating time is required; therefore, we used a high power to obtain accurately measurable temperature changes. The local SAR was calculated from the temperature rise using the following equation:

SAR (x, y ⫽ 0, z) ⫽ c ·

⌬T (x, y ⫽ 0, z) ⌬t

(1)

where c was the specific heat capacity of the muscle phantom material (3.63 kJ/kg/K), ⌬T was the local temperature rise in kelvins and ⌬t was the corresponding heating time in seconds. We assumed no electric field disturbing the effect of the fiber probes. IR thermography. We used IR thermography to visualize the temperature (rise) pattern at a high resolution (320 ⫻ 236 pixels). The pixel size was determined by taking an IR picture of a hot object with known dimensions. For our setup, the dimensions were 0.5 ⫻ 0.5 mm2. After the application of power, the top half was removed, and IR pictures were taken using a TVS-600 infrared camera (Nippon Avionics, Japan) that was mounted on a solid framework. From the IR picture before and after heating, we calculated the SAR pattern using Eq. 1. The inevitable time between the end of heating and the first IR picture was kept as short as possible (typically ⬃10 s) in an attempt to reduce the influences of heat conductivity.

Head-and-neck applicator prototype: verification and model

Fig. 4. Temperature (Tprobes) values and corresponding powers (PForw) and phases as function of time during example power pulse measurement. For fiber probe temperature measurements (visualized in dots), we created linear fits (solid lines) and calculated local dynamic SAR values from the slopes. Infrared pictures were captured before and after turning the power on and off.

Electric field Schottky-diode sheet. For fast measurements, which also enabled studying the prototypes’ dynamic behavior, we also performed SAR measurements with the electric field Schottky-diode sheet, as described by Kaatee and Van Rhoon (25) and Van Rhoon et al. (26). Then, two electric field sheets were positioned between both halves of the prototype (Fig. 3). To avoid a decrease in sensitivity caused by overlapping sheets at the location of the diodes (27), we positioned both sheets such that one row of measurement points was absent. The SAR values can be calculated from the amplitude of the measured electric field values, using the effective conductivity (␴eff) and mass density (␳), according to SAR (x, y ⫽ 0, z) ⫽ ␴ef f ·

ⱍE (x, y ⫽ 0, z)ⱍ 2␳

●

M. M. PAULIDES et al.

615

analyzer. Fiberoptic and IR measurements were conducted simultaneously. Hereto, we positioned an IR camera above the prototype and positioned the probes between both halves of the prototype. High power was applied, and we simultaneously measured the temperatures with the fiber probes (Fig. 4). After the required maximal temperature change was obtained, we opened the applicator and took IR pictures. After heating, we used a cooling time of typically 2 h between subsequent measurements to obtain a homogeneous thermal distribution. Electric field sheet measurements were performed separately at low power to avoid variations in the dielectric properties of the phantom from temperature differences. To check for measurement errors, all measurements were conducted multiple times within a 1-year period, and the muscle phantom was newly constructed two times. To obtain a central focus for each antenna (i) on each ring (j), we used central power (P) and phase (␸) settings (Pj,i ⫽ 25 W and ␸j,i ⫽ 0). Subsequently, we investigated the possibilities of SAR steering in radial and axial directions using a phase-steering method in analogy to the method used in the BSD2000 Sigma-60 system (28, 29). For each antenna, we analytically calculated the required phase delays (␸i,j), using the difference in wave velocity in muscle and water, such that the maximal interference occurs at the desired target center point (TCP). For this analytical model, we assumed a muscle cylinder (radius, 5.8 cm) and two circumferential arrays (radius, 14.15 cm) of six point-sources at z ⫽ ⫺4 cm and z ⫽ ⫹4 cm, and we neglected the PVC phantom shell.

RESULTS Reflection and cross-coupling After 1 year, the reflection coefficients had increased because of degradation (oxidation) of the setup (mainly the copper tape of the ground plane). Figure 5 shows the typical reflection coefficients of all antennas in the array at 23°C and 25°C halfway during our measurement period. Fig. 5 clearly shows that the reflection coefficient was influenced by the temperature of the water in the water bolus. For both temperatures, these reflection coefficients were all beneath ⫺10 dB (⬃6% reflection on average); thus, the efficiency of the antennas was sufficient for the purposes of our measure-

2

(2)

The electric field sheet directly measures, in real time, the axial (z) component of the electric field. Also, the variations in the dielectric properties of the phantom materials were much lower because measurements can be performed at low power (120 W). Furthermore, no manipulation of the setup is required for multiple measurements (i.e., all distributions can be measured using a single disposition of their experimental setup). Without calibration, their accuracy is 6% (26); however, when air gaps are present close to the diodes, this accuracy is reduced owing to the lower sensitivity. A main disadvantage of the present design of the sheet was its poor spatial resolution for measurements at 433 MHz (i.e., the distance between two diodes was 2.5 cm [⬃␭/4]). Measurement procedures and phase settings. As a first step, we measured the impedance characteristics of the antennas using an 8751A network analyzer (Agilent Technologies, Santa Clara, CA) to verify that they were sufficiently resonant. As a side step, we also assessed the amount of cross-coupling using the network

Fig. 5. Reflection coefficients at 433.92 MHz for each antenna of both arrays (e.g., 1\2 is the second antenna of the first antenna ring).

616

I. J. Radiation Oncology

●

Biology

●

Physics

(a)TCP = (−30,0,0) 1

FDTD T−probes IR E−sheet

0.75 0.5 0.25

Normalized SAR (W/kg)

0

−60

−40

−20

0

20

40

60

20

40

60

(b)TCP = (0,0,0) 1 0.75 0.5 0.25 0

−60

−40

−20

0

0.5

Water

0.75

Water

(c)TCP = (+30,0,0) 1

0.25 0

−60

−40

−20

0

20

40

60

x (mm)

Fig. 6. Simulated (finite-difference time-domain) and measured (temperature probes, infrared, and electric field sheet) dynamic SAR values at y ⫽ 0 mm and z ⫽ 0 mm for steering in x-direction. All tracks normalized to their respective maximal value.

ments. For the clinical applicator, however, we constructed a redesign that somewhat improved the reflection characteristics (17). For this qualitative analysis, we neglected the influences of the variations in efficiency (⬃6%) for the antennas of the prototype. We also found low cross-coupling between the antennas: maximum ⫺22 dB (0.6%) and, on average, ⫺27 dB (0.2%). The lowest values were obtained for neighboring antennas and the highest for antenna pairs with about one quadrant inner antenna spacing.

Volume 68, Number 2, 2007

influence of heat conductivity on the SAR measurements was negligible. Comparison measurements and simulation. Figure 6 shows the SAR values as function of the x-coordinate for TCP-based x-steering (i.e., for x ⫽ ⫺30 mm, x ⫽ 0 mm, and x ⫽ ⫹30 mm). Similarly, Fig. 7 shows the SAR values as function of the z-coordinate for TCP-based z-steering (i.e., for z ⫽ ⫺50 mm, z ⫽ ⫺25 mm, and z ⫽ 0 mm). For all IR thermography tracks, we averaged over 10 pixels (1 pixel was 0.7⫻0.7 mm2) to reduce noise. The tracks were normalized to their own maximum (IR thermography) or their own maximum of a spline-fit through the measurement points (temperature probes and electric field sheet measurements). For central steering, this maximum was typically ⬃0.6 W/kg (normalized to 1 W of total input power). The desired focus point (TCP) is always indicated with the dashed line. All measurements and predictions showed that a central focus was obtained. As shown by these values, this focus can be changed in the radial (x) and axial (z) directions. Comparing the results of the three measurement methods, we found a good qualitative agreement. A comparison of the simulations to these measurements revealed good agreement at the central positions, but discrepancies for more radial positions close to the PVC phantom border. The reason for this difference could be a result of inaccurate dielectric properties of the muscle phantom (lower conductivity) or air gaps at the connection between both halves of the setup. In the axial direction, an unexpected small asymmetry was present in the FDTD results for central settings (Fig. 7a). This was probably caused by the asymmetry in the feeding of the antennas (17).

(a)TCP = (0,0,0) 1

FDTD T−probes IR E−sheet

0.75 0.5 0.25 0

Normalized SAR (W/kg)

SAR tracks Figure 4 visualizes an example of the measured temperatures with the fiber probes and the corresponding measured powers and phases. It shows the length of the power pulse and the corresponding temperature rise from which we calculated the local SAR values. In this example case, the upper halve of the prototype was removed approximately 160 s after the power was turned on, and 10 s later, the IR picture was taken. The corrections by the control loop that were required to keep the power at the set value (25 W) are clearly visible in the power curves. At the start of the power pulse, the phases were influenced a little by the increase in power level, but these were automatically adjusted. Figure 4 also shows that two amplifiers were leaking some power after the power had been turned off, but this influence was relatively low (power on, 300 W vs. power off, ⬃5 W leakage). We found an almost perfect correlation between the measured temperatures and a linear fit, for all spatial locations (r2 ⫽ 1.0); thus, we concluded that the

−150

−100

−50

0

(b)TCP = (0,0,

50

100

150

100

150

100

150

−25)

1 0.75 0.5 0.25 0

−150

−100

−50

0

(c)TCP = (0,0,

50 −50)

1 0.75 0.5 0.25 0

−150

−100

−50

0

50

z (mm)

Fig. 7. Simulated (finite-difference time-domain) and measured (temperature probes, infrared, and electric field sheet) dynamic SAR values at x ⫽ 0 mm and y ⫽ 0 mm for steering in z-direction. All tracks normalized to their respective maximal value.

Head-and-neck applicator prototype: verification and model

Table 2. SAR steering (TCP setting and location of SARmax) and corresponding focus size (length and width of the 50% isoSAR contour) parameters for the simulated (FDTD) and measured (IR thermography) tracks TCP setting (mm)

Method

Central (0,0,0) X-steering (⫹30,0,0) Z-steering (0,0,⫺50)

FDTD IR FDTD IR FDTD IR

SARmax (mm)

L50% (mm)

W50% (mm)

(0,–,⫺1) (0,–,⫺5) (31,–,0) (32,–,0) (0,–,⫺50) (0,–,⫺51)

103 112 (⫹9%) 96 103 (⫹7%) 82 87 (⫹6%)

35 34 (⫺3%) 39 37 (⫺5%) 37 38 (⫹3%)

Abbreviations: SAR ⫽ specific absorption rate; TCP ⫽ target center point; SARmax⫽ maximal SAR; FDTD ⫽ finite-difference time-domain; IR ⫽ infrared; L50% ⫽ length of 50% iso-SAR; W50% ⫽ width of 50% iso-SAR.

●

M. M. PAULIDES et al.

617

imum was omitted. We indicated the location of the electric field sheet diodes by the horizontal dotted lines and the identity curve, indicating a perfect match, by the vertical dotted line. Both Figs. 8a and 8b show a linear relation and a high correlation between the TCP location and the location of the maximal SAR, in both radial (x, and y because of symmetry) and axial (z) directions. This meant that the focus could be accurately steered using the TCP method to the desired location in the neck-equivalent muscle phantom. We also found that steering in the radial plane was possible in the entire phantom. In the z-direction, the focus could be steered up to 60 mm from the center. From this location, a double focus distribution was obtained (i.e., the secondary focus became ⬎50% of the maximum).

DISCUSSION

X(SAR

X steering

50 Water

max

) (mm)

(a) 75 25 0 −25 −50

EFS(x,0,0) FDTD(x,0,0)

−75 −80

−60

−40

−20

0

X

TCP

Z(SAR

50 25 0

40

60

40

60

−25 −50 −75 −80

80

Z steering 2 maxima

max

) (mm)

(b)75

20

(mm)

2 maxima

Dynamic SAR steering We used measurements with the electric field sheet to investigate the dynamic steering possibilities of the focus, because no cooling down period is required for this method. We measured the location of maximal SAR and correlated this position to the TCP settings (Fig. 8). In Fig. 8, we show the maximal SAR location for x- and z-steering along several tracks. These curves showed a staircase-shape because of the spatial resolution of the electric field sheet, and because the use of a spline-fit to find the location of max-

Methods and materials For this work, we constructed a split-phantom setup consisting of two semicylinders that were closed with polymer sheets. In this way, the setup provided the versatility that enabled measurements between two parts with several measurement devices and IR measurements by opening the setup. The drawback of our approach was the possibility of air gaps between the sheets that, especially for these high frequencies, could have a large effect on the local SAR and on the measurement accuracy. We tried to decrease these influences by adding a little water between the sheets and

Water

In the radial direction, in which the SAR patterns are most steep, the IR thermography curves showed a somewhat flattened profile compared with the FDTD tracks. This flattening was caused by thermal conduction during the opening of the two halves after power application. After normalization, this resulted in relatively greater values in the valleys of the curves, because the maximal value was reduced. In the axial direction, this effect was less prominent, because the SAR curves were less steep. Furthermore, the increase in temperature was decreased by the circulated water of the water bolus of a stable temperature. The determination of the focus size (Table 2) provides a summary of the most important parameters from Fig. 6 and Fig. 7. Table 2 shows the SAR steering and corresponding focus size parameters for the simulated (FDTD) and measured (IR thermography) tracks. The simulated SAR distribution for central settings and the focus size parameters (L50%, W50%) were visualized in Fig. 3. For the experimental verification of the predicted focus size, we used the IR measurements because of the superior resolution and because the values correlated well with the temperature probe measurements, which we assumed to be more accurate. Table 2 shows that, using the TCP-based phases, the obtained SAR focus can be steered toward the desired location with reasonable accuracy (i.e., ⬃5 mm). Furthermore, it indicated that the corresponding focus lengths (L50%) were in the range of 87– 112 mm and the focus widths (W50%) were around 35 mm.

EFS(0,0,z) FDTD(0,0,z) −60

−40

−20

0

Z

TCP

20

80

(mm)

Fig. 8. Measured maximal dynamic-specific absorption rate point as function of target center point for dynamic-specific absorption rate steering in radial (a) and axial (b) directions. Locations of electric field sheet measurement points indicated by horizontal dotted lines and unity curve by diagonal dotted line. At z ⫽ ⫺25 mm, no electric fields were measured because of design of electric field sheets.

618

I. J. Radiation Oncology

●

Biology

●

Physics

some additional water pressure when the halves of the prototype were clamped together. For the measurements, we used the muscle phantom recipe of Ito et al. (18). The amounts of the ingredients were used for muscle-equivalent properties (⑀r ⫽ 56, ␴ ⫽ 0.8 S/m) at 22°C. These properties, however, vary with temperature and are very sensitive to small variations in salinity. Because of the required interference of multiple sources, these SAR values depend on the conductivity in a nonlinear fashion. By performing simulations with different conductivity values, we found that the SAR tracks in the x-direction were very dependent on conductivity (i.e., the lateral normalized SAR values decreased by 30% for a 20% decrease in conductivity). Therefore, in future measurements, we expect to improve the predictability of the measurements by measuring the properties of the phantom before each measurement. Results Electromagnetic models are used more and more for HT treatment planning, and the use of EM models to perform parameters studies for the optimization of HT applicators is growing (30 –32). Although the accuracy of EM modeling has increased dramatically and high-resolution simulations have become possible, to date, this has not led to higher standards concerning model validations. Such a validation is a crucial step to assess the dosimetry accuracy and will diminish the need for verification measurements for new applicators. Recently, for a superficial antenna, a first attempt of such a quantitative validation by high-resolution three-dimensional measurements was presented by de Bruijne et al. (22). This validation, however, was for a single antenna setup and should be confirmed for the SAR distributions with greater gradients that are normally obtained with array setups. For high-accuracy SAR dosimetry validations of array setups, accurate on-site measured dielectric properties of all (phantom) materials are required. Furthermore, an accurate measurement device is required, which, in general, is very expensive. For our approach, we used the availability of more accurate modeling to optimize the entire HT applicator. To our knowledge, we were the first to use EM modeling to design an HT applicator from scratch by detailed parameter studies at frequencies ⬎400 MHz. In a previous publication, we explained our work dedicated to the design of an optimized antenna positioning. Furthermore, we used EM modeling to fully optimize the reflection characteristics of a single element of the antenna array. In the present study, we have experimentally verified the feasibility of focused heating in the neck, which was predicted earlier (12, 14), and to validate our approach. We also assessed the dimensions of the focus and verified the possibilities of dynamic steering. By comparing the measurements to simulation, we performed a qualitative verification of the predicted SAR patterns in a muscle phantom. In this study, we verified that a central focus can be obtained with a double ring of six antennas each. For central

Volume 68, Number 2, 2007

settings, we determined the width of this focus as 34 mm (IR) and 35 mm (FDTD) and the length as 112 mm (IR) and 103 mm (FDTD). This difference might be caused by thermal conduction, which resulted in an overestimation of the effective field size (EFS) by IR thermography. In the radial direction, this size is normally ␭/3 (13, 27), which results in ⬃31 mm in a muscle phantom, and correlated well with the values in our study. In the axial direction, this focus was determined by a combined effect of the distance between the antenna rings, the radius of the antenna rings, and the operating frequency (i.e., the wavelength in muscle) (14). The obtained focus could be steered in radial directions toward every point in the muscle phantom. In the axial direction (⫺z,z), the extreme focus center point location was a 60-mm off-center position. For verification of the SAR profiles, we used the splitphantom technique in combination with glass-fiber thermometer probe measurements, IR thermography using a high-resolution IR camera, and measurements using our electric field Schottky-diode sheet. The local SAR values were determined for the first two (fiber probes and IR) using the power-pulse technique (33). For these, high power is required; the latter (electric field) measurements were done at low power. The fiber probe measurements have the advantage of high accuracy, and they provide the possibility of monitoring temperatures during heating. Furthermore, they are nonmetallic and very thin; thus, the field disturbances were negligible. The disadvantage of this method is the required cooling time after measurement (1–2 h) because of the high temperature gradients required. Furthermore, multiple (multisensor) probes are required for high resolution, and positioning multiple probes in the split-phantom was cumbersome. IR thermography measurements using the most recent IR cameras have a high resolution that enables the investigation of not only the size of the focus, but also the shape of the focus. IR measurements share the disadvantage of long cooling times. Furthermore, for these measurements, opening the setup is required to be able to take an IR picture (i.e., measurements were taken typically 10 s after the power was off). According to Samaras et al. (24), for EFS measurements, this is sufficiently short for an accurate measurement. Although, compared with flat-phantom EFS measurements, the distributions obtained in this publication showed much greater SAR gradients and, therefore, heat conduction might have somewhat more influence, by comparing the thermography results with those of the fiber probe measurement, we could show that the influences of heat conduction remained low. Using the electric field Schottkydiode sheet (25, 26), we performed electric field measurements that can be performed at low power; thus, dynamic SAR steering with the prototype could be assessed. A major drawback of this equipment is its low resolution, because it was designed for superficial HT applicator assessments in flat phantoms in which the fields are more gradual. Furthermore, in this split-phantom setup, air gaps may be present that considerably decrease the sensitivity of the diodes. The superficial SAR values measurements by both power-pulse– based methods

Head-and-neck applicator prototype: verification and model

●

M. M. PAULIDES et al.

619

also might have been hampered by the active water cooling of the sides of the phantom. From the resemblance between the SAR derived from the electric field and the temperature measurements, we concluded that this influence was low.

homogeneous temperature distribution, an improvement would be to use a liquid phantom material. The required new electric field sheet system, however, must be specifically designed and other questions remain to be solved.

Quality assurance Currently, no specific quality assurance guidelines have been defined for our application or for newly developed applicators. Within the framework of the European Society for Hyperthermic Oncology (ESHO), guidelines for ESHO protocols have been defined: (1) for treatment of superficial tumors (33), including locally advanced breast carcinoma, advanced neck lymph nodes, malignant melanoma; and (2) for regional HT in the pelvic area by radiofrequency equipment (34). For regular assessment of the SAR pattern of an applicator, different approaches are suggested. For superficial quality assurance guidelines and the assessment of applicator performance, the EFS and penetration depth are defined, and “should be determined by measuring the changes in temperature resulting from a brief pulse of high power” (33). In these guidelines, the necessity for an additional characterization of the SAR pattern corresponding to a multielement array of applicators is recognized. Furthermore, if “nonperturbing electric field probes, thermographic imaging or liquid-crystal sheet imaging” are used, they should be corroborated by measurements obtained using the power pulse technique. A requirement for using this technique is that 60 s after the power is turned on, the EFS and penetration depth should be determined. However, by finding perfect linear temperature curves, we showed that longer heating periods can be used. In the guidelines for regional HT (34), characterization of the equipment to perform system performance regularly using the LED matrix/lamp phantom using different phase/amplitude settings, different phantom positions, and different bolus configurations is required for SAR distribution. The electric field Schottky-diode sheet system is compliant with all requirements, but a greater resolution is required; thus, the focus size and shape can also be determined (i.e., ⬃␭/10 instead of ⬃␭/4 at 433 MHz). Also, to obtain a more

CONCLUSION In this study, we analyzed the ability of using a phasedarray HT applicator for focused heating in the head-andneck region. First, we have concluded that the early design of the patch antenna, in a well-controlled water environment, provided an average efficiency of 94%, which was highly sufficient for our investigations. Furthermore, we measured low cross-coupling between antennas in the array (maximum, ⫺22 dB). Third, we found that, using the analyzed setup of two rings of six antennas, focused heating in the head and neck is feasible. Using IR thermography, we measured a central 50% iso-SAR focus of 112 mm in the axial and 34 mm in the radial directions in a muscle phantom. This radial dimension was sufficiently small to allow targeted heating of advanced tumors, and, by defocusing, good SAR coverage of larger tumors should also be possible. The axial length of this focus was somewhat large for tumor-specific heating; however, we expect that this results in blood preheating, because the vessels are mainly oriented in this direction. The focus can be steered in axial direction to maximum 60 cm from the center in the phantom. Combined with the focus size of 10 cm in the axial direction, this indicates that 50% SAR coverage could be possible ⱕ11 cm from a central location. The actual extent of heating in a patient remains to be determined. We found little differences between the predictions of the SAR distribution using the three investigated measurement methods. Therefore, as the best system for regular quality assurance, we chose the electric field Schottky-diode sheet system. This device enables quick measurements and better control of the temperatures, resulting in more predictable dielectric properties of the phantom materials. However, a modified design of the electric field Schottky-diode sheet system is required to provide a greater resolution and better reduction possibilities of air gaps.

REFERENCES 1. El-Sayed S, Nelson N. Adjuvant and adjunctive chemotherapy in the management of squamous cell carcinoma of the head and neck region: A meta-analysis of prospective and randomized trials. J Clin Oncol 1996;14:838 – 847. 2. Pignon JP, Bourhis J, Domenge C, et al. Chemotherapy added to locoregional treatment for head and neck squamous-cell carcinoma: Three meta-analyses of updated individual data. Lancet 2000;355:949 –955. 3. Trotti A. Toxicity in head and neck cancer: A review of trends and issues. Int J Radiat Oncol Biol Phys 2000;47:1–12. 4. Huilgol NG. A phase I study to study arsenic trioxide with radiation and hyperthermia in advanced head and neck cancer. Int J Hyperthermia 2006;22:391–397. 5. van der Zee J, Gonzalez Gonzalez D, van Rhoon GC, et al., for the Dutch Deep Hyperthermia Group. Comparison of radiotherapy alone with radiotherapy plus hyperthermia in

6.

7. 8. 9.

locally advanced pelvic tumours: A prospective, randomised, multicentre trial. Lancet 2000;355:1119 –1125. Colombo R, Da Pozzo LF, Salonia A, et al. Multicentric study comparing intravesical chemotherapy alone and with local microwave hyperthermia for prophylaxis of recurrence of superficial transitional cell carcinoma. J Clin Oncol 2003;21:4270– 4276. Jones E, Thrall D, Dewhirst MW, et al. Prospective thermal dosimetry: The key to hyperthermia’s future. Int J Hyperthermia 2006;22:247–253. Valdagni R, Liu F-F, Kapp S. Important prognostic factors influencing outcome of combined radiation and hyperthermia. Int J Radiat Oncol Biol Phys 1988;15:959 –972. Valdagni R, Amichetti M. Report of long-term follow-up in a randomized trial comparing radiation therapy and radiation therapy plus hyperthermia to metastatic lymph nodes in stage

620

10. 11. 12.

13.

14. 15.

16. 17. 18. 19. 20. 21. 22.

I. J. Radiation Oncology

●

Biology

●

Physics

IV head and neck patients. Int J Radiat Oncol Biol Phys 1993;28:163–169. Wust P, Seebass M, Nadobny J, et al. Simulation studies promote technological development of radiofrequency phased array hyperthermia. Int J Hyperthermia 1996;12:477– 494. Stauffer PR. Evolving technology for thermal therapy of cancer. Int J Hyperthermia 2005;21:731–744. Paulides MM, Vossen SHJA, Zwamborn APM, et al. Theoretical investigation into the feasibility to deposit RF energy centrally in the head and neck region. Int J Radiat Oncol Biol Phys 2005;63:634 – 642. Paulides MM, Wielheesen DHM, van der Zee J, et al. Assessment of the local SAR distortion by major anatomical structures in a cylindrical neck phantom. Int J Hyperthermia 2005; 21:125–140. Paulides MM, Bakker JF, Zwamborn APM, et al. A head and neck hyperthermia applicator: Theoretical antenna array design. Int J Hyperthermia 2007;23:59 – 67. Seebass M, Beck R, Gellerman J, et al. Electromagnetic phased arrays for regional hyperthermia: Optimal frequency and antenna arrangement. Int J Hyperthermia 2001;17:321– 336. Paulsen KD, Geimer S, Tang J, et al. Optimisation of pelvic heating rate distributions with electromagnetic phased arrays. Int J Hyperthermia 1999;15:157–186. Paulides MM, Bakker JF, Chavannes N, et al. A patch antenna design for a phased-array head and neck hyperthermia applicator. IEEE Trans Biomed Eng. 2007; In press. Ito K, Furuya K, Okano Y, et al. Development and characteristics of a biological tissue-equivalent phantom for microwaves. Electron Commun Jpn 2001;84:67–77. Gabriel S, Lau RW, Gabriel C. The dielectric properties of biological tissues III: Parametric models for the dielectric spectrum of tissues. Phys Med Biol 1996;41:2271–2293. SEMCAD: The simulation platform for EMC, antenna design and dosimetry. Smith and Partner Engineering AG. Available at: www.semcad.com. Accessed January 2, 2006. Samaras T, Rietveld PJM, van Rhoon GC. Effectiveness of FDTD in predicting SAR distributions from the Lucite cone applicator. IEEE Trans Micr Theor Tech 2000;48:2059 –2063. de Bruijne M, Samaras T, Neufeld E, et al. Quantitative three-dimensional SAR validation of the Lucite cone applica-

Volume 68, Number 2, 2007

23.

24.

25.

26.

27.

28. 29.

30.

31.

32.

33.

34.

tor. Abstracts of 22nd Annual Meeting of the SHO, Graz, Austria, June 2005. Stogryn A. Equations for calculating the dielectric constant of saline water (correspondence). IEEE Trans Micr Theor Tech 1971;19:733–736. Samaras T, van Rhoon GC, Sahalos JN. Theoretical investigation of measurement procedures for the quality assurance of superficial hyperthermia applicators. Int J Hyperthermia 2002; 18:416 – 425. Kaatee RSJP, van Rhoon GC. An electric field measurement system, using a two-dimensional array of diodes. Int J Hyperthermia 1999;15:441– 454. Van Rhoon GC, Ameziane A, Lee WM, et al. Accuracy of electrical field measurement using the flexible Schottky diode sheet at 433 MHz. Int J Hyperthermia 2003;19:134 –144. Van Rhoon GC, van der Heuvel DJ, Ameziane A, et al. Characterization of the SAR-distribution of the Sigma-60 applicator for regional hyperthermia using a Schottky diode sheet. Int J Hyperthermia 2003;19:642– 654. Sigma eye applicator operator manual. BSD Medical Corporation, Salt Lake City, UT; 1995. Turner PF, Schaeffermeyer T. BSD-2000 approach for deep local and regional hyperthermia. Strahlenther Oncol 1989; 165:738 –741. Kroeze H, van de Kamer JB, De Leeuw AA, et al. Regional hyperthermia applicator design using FDTD modelling. Phys Med Biol 2001;46:1919 –1935. Nadobny J, Wlodarczyk W, Westhoff L, et al. Development and evaluation of a three-dimensional hyperthermia applicator with Water-COated Antennas (WACOA). IEEE Med Phys 2003;30:2052–2064. Siauve N, Nicolas L, Vollaire C, et al. Optimization of the sources in local hyperthermia using a combined finite elementgenetic algorithm method. Int J Hyperthermia 2004;20:815– 833. Hand JW, Lagendijk JJW, Bach Andersen J, et al. Quality assurance guidelines for ESHO protocols. Int J Hyperthermia 1989;5:421– 428. Lagendijk JJW, Van Rhoon GC, Hornsleth SN, et al. ESHO quality assurance guidelines for regional hyperthermia. Int J Hyperthermia 1998;14:125–133.