2013 First International Conference on Artificial Intelligence, Modelling & Simulation
Genetic Algorithm Optimization of SAR Distribution in Hyperthermia Treatment of Human Head Maged Aldhaeebi, Mohammad Alzabidi and Ibrahim Elshafiey Department of Electrical Engineering King Saud University Riyadh, Saudi Arabia
[email protected],
[email protected], and
[email protected]
phased array elements, corresponding to each frequency of excitation sub-channels. A module in SEMCAD-X for optimizing electromagnetic field and SAR distribution is used in this research [8]. Excitation is implemented using an array of Vivaldi antenna elements. UWB Vivaldi antenna presents an attractive configuration for hyperthermia and microwave imaging [9-11]. Fig. 1 shows an illustration of the antenna array surrounding a human head phantom. The distance between the phantom and the antenna is set to one mm. Water background is used to allow reduction in antenna size, enhance the matching with the human tissue and achieve cooling of the elements.
Abstract—Optimization is performed of the specific absorption rate (SAR) distribution inside a human head phantom for hyperthermia treatment of brain tumor. Two implementations of phased array are presented, using four and eight antenna, respectively. SEMCAD-X simulation along with genetic algorithm (GA) optimization under Matlab environment are conducted. Operation is performed at relatively low frequency values of 550-900 MHz to achieve good energy penetration. Results of SAR distribution illustrates the possibility of delivering microwave energy into relatively deep tumor positions with reduced heating side effects on the surrounding healthy tissue. Keywords- Hyperthermia treatment; Genetic algorithm (GA), Optimization, specific energy absorption rate (SAR).
I. INTRODUCTION Cancer is a challenging disease that accounts for around 13% of all deaths [1]. Various treatment plans are clinically used. Hyperthermia is an arising plan that is used to improve the capabilities of cancer treatment using temperature increase in the tissue in range 40–44 °C [2]. Three types of hyperthermia: whole-body, regional, and local hyperthermia [3]. Depending on the tumor location, there various methods are used to apply local hyperthermia, including external, intra-luminal and interstitial methods [4]. Non-invasive delivery of electromagnetic energy into deep tumors while reducing the side effects on healthy tissues is a challenge that hampers the development of hyperthermia treatment [5] and [6]. Recent technology advances, particularly the maturity of ultra-wideband (UWB) systems, and the progress in phased-array tools help to enhance the use of external applicators in hyperthermia treatment [7]. UWB external phased array applicators provide an appealing solution for the development of hyperthermia treatment. The progress in optimization tools is also useful to improve the performance of external hyperthermia applicators, by achieving designs to focus the microwave energy into the tumor location and decrease the hot spots in healthy tissue. This paper investigates the optimization of the distribution of specific the absorption rate (SAR) in the intended tumor location for hyperthermia treatment of brain tumor depending on genetic algorithms optimization. The objective is to find optimum phase and amplitude values for
978-1-4799-3251-1/13 $31.00 © 2013 IEEE DOI 10.1109/AIMS.2013.22
Figure 1. Illustration of the proposed Configuration.
II. SPECIFIC ABSORPTION RATE (SAR) SAR is a quantity that describes the amount of absorbed radiated effect for a specific material at a certain frequency [12]. The specific absorption rate (SAR) is calculated as =
=
||
(1)
Where P is the power absorbed in the tissue in W; is the mass density in kg/ m3, is the electrical conductivity in S/m, and |E| is the rms magnitude of the electric field strength vector in V/ m [13]. SAR distribution induced by the external heating device should be determined by Pennes’ mathematical model [14, 15]. III.
PHASED ARRAYS TECHNIQUES
The phased array concept is based on the use of individual elements that can be driven independently to control the relative phases of the feeding signals. The use of 76 82 92
preserved from population to population, to avoid the loss of good solutions along the optimization. SAR distribution corresponding to four and eight antenna array before optimization is shown in Fig. 4 to Fig. 11. The figures are shown for a cut-plane at vertical distance z=0. The horizontal axis in the figures points in the x-direction, while the other normal direction is the y-axis. The Considered frequency values are 250, 500, 750 and 900 MHz, respectively. All magnitude values are set to one and the phases are set to zero. SAR values are calculated corresponding to input power of 500 W. Average is performed per 1-g of tissue. The figures reveal the effect of frequency value on energy penetration in the tissue.
a phase array offers the opportunity to achieve constructive interfere of the electromagnetic energy at the desired locations, while minimizing the energy in healthy and crucial tissues, based on destructive interference [16]. Circular phased array antennas are appealing in many applications since they show uniform beam coverage over 360 degrees in, and they can generate radiation patterns with a main beam [17]. The brain phantom model is taken to be a cylinder with a radius of 84 mm with a dielectric of electrical parameters equal to those of the brain tissue, such that the relative permittivity is 40, the electrical conductivity is 1 S/m and the mass density of the phantom material is 1000 kg/m3. A small sphere is used to represent the tumor with a radius of 10 mm. The relative permittivity is taken to be 38, and the electrical conductivity is 0.5 S/m. Two circular array configurations are used with four and eight elements, respectively. The elements surround the phantom, while they are immersed in water background. The antennas are set at 1 mm from the phantom as show in Fig.2.
Population Initialization
Fitness Computation
Yes Is the Objective Achieved
End
No Selection
N Times Crossover
Mutation Figure 2. Eight-element phased array.
IV. SIMULATION RESULTS The simulation setup is based on SEMCAD-X environment that implements three-demission finite difference time-domain (FDTD) modeling. SEMCAD-X environment provides fast simulation capabilities, computing up to 400-1500 Mcells/s with the use of the hardware acceleration. In this analysis, we investigate the effect of phase and amplitude on the specific absorption rate distribution in tumor location. Genetic algorithm based optimization is used due to their capability in finding the global optima in complex optimization problems, such as electromagnetic field problems [18, 19]. The optimization module in SEMCAD-X allows the user to perform the optimization with arbitrary number of parameters using combinations of goal functions with adjustable weights. The general GA flow-chart is shown in Fig. 3. A specific percentage of the best chromosomes is
Mirror Figure 3. Genetic algorithm flow chart in SEMCAD-X tool.
Figure 4. SAR distribution of four antenna array without optimization at frequency of 250 MHz. Scale is in dB.
93 83 77
Figure 5. SAR distribution of four antenna array without optimization at 500 MHz. Scale is in dB. Figure 9. SAR distribution of eight antenna array without optimization at 500 MHz. Scale is in dB.
Figure 6. SAR distribution of four antenna array without optimization at 750 MHz. Scale is in dB. Figure 10. SAR distribution of eight antenna array without optimization at 750 MHz. Scale is in dB.
Figure 7. SAR distribution of four antenna array without optimization at 900 MHz. Scale is in dB.
Figure 11. SAR distribution of eight antenna array without optimization at 900MHz. Scale is in dB.
Performing GA based optimization, the optimum amplitude and phase values of each antenna element from the array are identified. The objective is set to maximize the SAR values at the tumor location, while minimizing the values in other brain tissues. Fig. 12 and Fig. 13 present the results of SAR at frequency of 500 MHz for both four and eight element phased array implementation, respectively. Fig. 14 and Fig. 15 show the spatial peak SAR of both four and eight element array, respectively.
Figure 8. SAR distribution of eight antenna array without optimization at 250 MHz. Scale is in dB.
94 84 78
Figure 12. SAR distrbuation with optimization using four antenna array at 500 MHz.
Figure 15. Spatial peak SAR using eight antenna array at 750 MHz.
Fig. 16 and Fig. 17 present SAR distribution inside the phantom, after optimization as a function of y-distance, for x= 20 mm and z =0 mm. The two figures correspond to four-element and eight-element arrays, respectively.
Figure 13. SAR distrbuation with optimization using eight antenna array at 500 MHz. Figure 16. SAR distribution profile as a function of y, using four element array. x= 20 mm and z =0.
Figure 14. Spatial peak SAR using four antenna array at 750 MHz Figure 17. SAR distribution profile as a function of y, using eight element array. x= 20 mm and z =0.
95 85 79
A genetic algorithm based optimization tool is also developed under MATLAB optimization environment [20]. A flow-chart of the tool is shown in Fig. 18. Simulation is performed considering a phantom model with one positioned antenna. The tumor is rotated by 45 degree; starting from the original position zero to 315 degrees. The simulated electrical field maps are exported to the tool to combine the electrical field of all experiments. The tool is used to identify the optimum amplitude and phase values. The following parameters are used: population size of 100, and the number of iterations of 100. Fig. 19 and Fig. 20 show SAR distribution before conducting optimization for frequency values of 500MHz and 750MHz, respectively. The magnitudes values are set to one for each antenna and the phase values are set to zero. The result of optimum SAR distribution is show in Fig. 21 and Fig. 22 for frequency values of 500MHz and 750MHz, respectively. The figures show that the SAR distribution is directed towards the tumor location.
Figure 19. SAR distribution without optimization at 500 MHz.
Conduct Simulation
Rotate Tumor by 45o
Export Field to Tool
Interpolated the Results (1mm step) Figure 20. SAR distribution without optimization at 750 MHz
Rotate Results (tumor back to original position)
Add Fields (Each multiplied by a factor)
Using GA technique to find optimum phase & amplitude
Plot SAR
End
Figure 21. SAR distribution after performing GA optimization at 500 MHz.
. Figure 18. Flow chart for developed optimization tool.
96 86 80
[3]
[4]
[5]
[6]
[7]
[8] Figure 22. SAR distribution after performing GA optimization at 750 MHz.
[9]
V. CONCLUSIONS In this paper, optimization of excitation of external applicators for hyperthermia treatment of brain is considered The specific energy absorption rate (SAR) distribution is investigated inside phantom model. Two tools are developed. The first depend on using SEMCAD-X optimization module. The second tool is developed using MATLAB genetic algorithm environment. The objective is set to find the optimum amplitude and phase of each array elements to maximize SAR value and enhance the localization of the microwave energy inside the tumor location. Simulation has been conducted at various subchannels of relatively low frequency values of 550MHz, 700MHz and 900MHz.
[10]
[11]
[12]
[13]
[14]
ACKNOWLEDGMENT [15]
This research work is supported by the National Plan for Science and Technology (NPST), Kingdom of Saudi Arabia, under project number: 10-ELE996-02.
[16]
REFERENCES [1]
[2]
[17]
World Health Organization. (Reviewed January 2013). Cancer: Fact sheet. Available: http://www.who.int/mediacentre/factsheets/fs297/en/ N. S. Karnik, R. Tulpule, M. Shah, P. S. Verma, C. Y. Huang, J. Y. Cha, A. Pandya, S. Usman, V. Pulipati, P. Pagadala, and B. P. Kumar, "Design, Simulation and Experimental Study of Near-Field Beam forming Techniques Using Conformal Waveguide Arrays," Microwaves, Antennas & Propagation, IET, vol. 4, pp. 162-174, 2010.
[18] [19] [20]
97 87 81
D. K. Chatterjee, P. Diagaradjane, and S. Krishnan. (2011) Nanoparticle-Mediated Hyperthermia in Cancer Therapy. Therapeutic Delivery. 1001-14. M. Converse, E. J. Bond, B. D. V. Veen, and S. C. Hagness, "A Computational Study of Ultra-Wideband Versus Narrowband Microwave Hyperthermia for Breast Cancer Treatment," IEEE Transactions on Microwave Theory and Techniques, vol. 54, pp. 2169-2180, 2006. A. L. E. Rast, "Examination Of Ideal Physical Parameter Choices For Efficacious Hyperthermic Treatment Of Cancer," PhD Dissertation, University of Alabama at Birmingham, USA, 2011. M. E. Kowalski, "Optimization of Electromagnetic Phased-Arrays for Hyperthermia via Magnetic Resonance Temperature Estimation," IEEE Transactions on Biomedical Engineering, vol. 49, pp. 12291241, 2002. D. Correia, "Body Conformal Antennas for Superficial Hyperthermia: The Impact of Bending Contact Flexible Microstrip Applicators on their Electromagnetic Behavior," IEEE Transactions on Biomedical Engineering, vol. 56, pp. 2917-2926, 2009. Speag. (2013). SEMCAD X. Available: http://www.speag.com/products/semcad/overview/ M. Aldhaeebi, M. Alzabidi, and I. Elshafiey, "Optimization of UWB Antenna Array for Hyperthermia Treatment of Brain Tumor," in Electronics, Communications and Photonics Conference (SIECPC), 2013 Saudi International, 2013, pp. 1-6. M. A. Alzabidi, M. A. Aldhaeebi, and I. Elshafiey, "Development of UWB Vivaldi Antenna for Microwave Imaging," in Electronics, Communications and Photonics Conference (SIECPC), 2013 Saudi International, 2013, pp. 1-4. I. Elshafiey, A. F. Sheta, Z. Siddiqui, and M. A. Alkanhal, "UltraWideband Applicator for Brain-Tumor Ablation and Imaging System," in Signal Processing and Information Technology (ISSPIT), 2012 IEEE International Symposium on, 2012, pp. 000310-000314. N. Aronsson and D. Askeroth, "A Comparative Study of Electromagnetic Dosimetric Simulations and Measurements," MS Thesis, Department of Electroscience, Lund university, April 2002. S. Curto, "Antenna Development for Radio Frequency Hyperthermia Applications," MS Thesis, Dublin Institute of Technolgy School of Electronic & Communications Engineering, 2010. K. D. Paulsen, "Calculation of power deposition patterns in hyperthermia," in In Thermal Dosimetry and Treatment Planning,M Gautherie, Ed. , Berlin, Germany:. H. H. Pennes, "Analysis of Tissue and Arterial Blood Temperatures in the Resting Human Forearm," Journal. Applied. Physiolog, vol. 1, pp. 93-122, 1984. J. Paguet, A. Garcia, J. Vazquez, J. Marquet, and F. Pichonnat, " Phased Array technology concepts,probes and applications," in 8th European Congress on Non Destructive testing, Barcelona - Spain, 2002, pp. 26-54. H. Steyskal, "Digital Beamforming Aspects of Wideband Circular Arrays," in Aerospace Conference, Big Sky, MT, 2008. R. L. Haupt and D. H. Werner, Genetic Algorithms in Electromagnetics: Wiley. com, 2007. Y. Rahmat-Samii and E. Michielssen, Electromagnetic Optimization by Genetic Algorithms: John Wiley & Sons, Inc., 1999. MathWorks. Optimization Toolbbox: Solve Standard and LargeScale Optimization Problems. Available: http://www.mathworks.com/products/optimization/
