Bir Çizgi Kaynağı Tarafından Oluşturulan İnsan Meme Modelinde 2B Elektrik Alan Dağılımının Yapay Zeka Destekli Tahmini

Abstract

Bu çalışma, mikrodalga görüntüleme ve hipertermi gibi tıbbi uygulamalarda kritik bir görev olan heterojen meme dokusu modelleri içindeki iki boyutlu elektrik alan (EF) dağılımlarını tahmin etmenin hesaplama zorluklarını ele almaktadır. Geleneksel sayısal simülasyon yöntemleri doğrudur ancak yüksek hesaplama yükü getirir ve genellikle dakikalarca süren işlem süresi gerektirir. Bu sınırlamanın üstesinden gelmek için, EF dağılımlarını doğrudan dielektrik özellik haritalarından, özellikle elektrik iletkenliği ve geçirgenlikten tahmin eden ve çıkarım süresini önemli ölçüde indiren bir derin öğrenme yaklaşımı öneriyoruz. Bu çalışmada iki evrişimsel sinir ağı mimarisi değerlendirilmiştir: ResNeXt50 kodlayıcılı bir U-Net modeli (R50-UNet) ve standart bir ResNeXt50 modeli (R50). Ek olarak, etki alanı içinde en yüksek alaka düzeyine sahip bölgelerde öğrenmeyi vurgulayan maskeleme tabanlı bir kayıp fonksiyonu sunulmuştur. Nicel değerlendirme, R50-UNet modelinin standart R50 modelinden daha iyi performans gösterdiğini ve sinyal-gürültü oranında (SNR) 14,66 dB'ye kadar iyileşme sağladığını göstermektedir. Maskeleme yönteminin uygulanması, SNR'de 3,24 dB'ye varan ek bir kazançla performansı daha da artırmaktadır. Veri verimliliği analizi, R50 modelinin mevcut eğitim verilerinin yalnızca %25'i ile performans doygunluğuna ulaşırken, maskelenmiş kayıp ile birleştirilmiş R50-UNet mimarisinin daha fazla veri kullanıldıkça iyileşmeye devam ettiğini ortaya koymaktadır. Bu bulgular, hesaplama süresinin ve uzamsal hassasiyetin kritik olduğu biyomedikal senaryolarda hızlı ve doğru EF tahmini için derin öğrenme kullanımının uygulanabilirliğini desteklemektedir.

Keywords

• derin öğrenme
• elektrik alan dağılımı tahmini
• dielektrik özellikler
• mirkrodalga hipertermi
• mikrodalga görüntüleme,

Supporting Institution

Scientific and Technological Research Council of Türkiye (TUBITAK)

AI-Driven 2D Electric Field Distribution Estimation in a Human Breast Model Formed by a Line Source

Abstract

This study addresses the computational challenges of estimating two-dimensional electric field (EF) distributions within heterogeneous breast tissue models, a critical task in medical applications such as microwave imaging and hyperthermia. Traditional numerical simulation methods are accurate but computationally expensive, often requiring minutes of processing time. To overcome this limitation, we propose a deep learning approach that predicts EF distributions directly from dielectric property maps, specifically electrical conductivity and permittivity, significantly reducing inference time to a matter of seconds. Two convolutional neural network architectures are evaluated: a U-Net model with a ResNeXt50 encoder (R50-UNet) and a standard ResNeXt50 model (R50). Additionally, we introduce a masking-based loss function that emphasizes learning in regions of highest relevance within the domain. Quantitative evaluation demonstrates that the R50-UNet model outperforms the standard R50 model, achieving up to a 14.66 dB improvement in signal-to-noise ratio (SNR). The application of the masking method further enhances performance, with an additional gain of up to 3.24 dB in SNR. Data efficiency analysis reveals that while the R50 model reaches performance saturation with only 25% of the available training data, the R50-UNet architecture combined with masked loss continues to improve as more data are utilized. These findings support the feasibility of using deep learning for fast and accurate EF prediction in biomedical scenarios where computation time and spatial precision are critical.

Keywords

• deep learning
• electric field distribution estimation
• dielectric properties
• microwave hyperthermia
• microwave imaging

Supporting Institution

Scientific and Technological Research Council of Türkiye (TUBITAK)

Ethical Statement

This study does not require an ethical approval. The utilized data are available online and the corresponding references were given.

Thanks

This study was supported by Scientific and Technological Research Council of Türkiye (TUBITAK) under the Grant Number 125E412. This work was also supported by Research Fund of the Istanbul Technical University (ITU) with Project Number MAB-2023-44842. The authors thank to TUBITAK and ITU for their supports.

References

1. [1] A. Janjic, I. Akduman, M. Cayoren, O. Bugdayci, and M. Aribal, “Safe—microwave imaging device for breast cancer early screening and diagnostics,” in Electromagnetic Imaging for a Novel Generation of Medical Devices: Fundamental Issues, Methodological Challenges and Practical Implementation, 2023, pp. 273–292.
2. [2] E. Maheshwari et al., “Update on MRI in evaluation and treatment of endometrial cancer,” Radiographics, vol. 42, pp. 2112–2130, 2022.
3. [3] G. Altintas, I. Akduman, A. Janjic, and T. Yilmaz, “A novel approach on microwave hyperthermia,” Diagnostics, vol. 11, p. 493, 2021.
4. [4] H. Kasban, M. El-Bendary, and D. Salama, “A comparative study of medical imaging techniques,” Int. J. Inf. Sci. Intell. Syst., vol. 4, pp. 37–58, 2015.
5. [5] B. Acar, T. Y. Abdolsaheb, and A. Yapar, “Advanced hyperthermia treatment: optimizing microwave energy focus for breast cancer therapy,” Turk. J. Electr. Eng. Comput. Sci., vol. 32, pp. 268–284, 2024.
6. [6] A. Turgut and B. Engiz, “Analyzing the SAR in human head tissues under different exposure scenarios,” Appl. Sci., vol. 13, p. 6971, 2023.
7. [7] A. Hirata et al., “Assessment of human exposure to electromagnetic fields: Review and future directions,” IEEE Trans. Electromagn. Compat., vol. 63, pp. 1619–1630, 2021.
8. [8] N. Iacob, “Pitfalls and challenges in specific absorption rate evaluation for functionalized and coated magnetic nanoparticles used in magnetic fluid hyperthermia,” Coatings, vol. 15, p. 345, 2025.

9. [9] F. Kahnert, “Numerical methods in electromagnetic scattering theory,” J. Quant. Spectrosc. Radiat. Transfer, vol. 79, pp. 775–824, 2003.
10. [10] M. N. Sadiku, Numerical Techniques in Electromagnetics with Matlab, 3rd ed. Boca Raton, FL, USA: CRC Press, 2012.
11. [11] J. Jin, The Finite Element Method in Electromagnetics, 3rd ed. Hoboken, NJ, USA: Wiley, 2015.
12. [12] R. Harrington, Field Computation by Moment Methods, New York, NY, USA: IEEE Press, 1993.
13. [13] S. Joof et al., “A guideline for complex permittivity retrieval of tissue-mimicking phantoms from open-ended coaxial probe response with deep learning,” IEEE Trans. Microw. Theory Techn., vol. 70, pp. 5105–5115, 2022.
14. [14] C. Aydinalp, S. Joof, M. Akinci, I. Akduman, and T. Yilmaz, “Microwave dielectric property retrieval from open-ended coaxial probe response with deep learning,” IEEE Access, vol. 10, pp. 1216–1227, 2021.
15. [15] P. Wiecha and O. Muskens, “Deep learning meets nanophotonics: A generalized accurate predictor for near fields and far fields of arbitrary 3D nanostructures,” Nano Lett., vol. 20, pp. 329–338, 2019.
16. [16] M. Li and Z. Ma, “Accurate prediction of electric fields of nanoparticles with deep learning methods,” IEEE J. Multiscale Multiphysics Comput. Tech., vol. 8, pp. 178–186, 2023.
17. [17] H. M. Yao, L. Jiang, and M. Ng, “Implementing the fast full-wave electromagnetic forward solver using the deep convolutional encoder-decoder architecture,” IEEE Trans. Antennas Propag., vol. 71, no. 1, pp. 1152–1157, 2022.
18. [18] H. M. Yao, L. Jiang, and M. Ng, “Enhanced deep learning approach for electromagnetic forward modelling of dielectric target within the wide frequency band using deep residual convolutional neural network,” IEEE Antennas Wireless Propag. Lett., vol. 23, no. 6, pp. 1884–1888, 2024.
19. [19] S. Ayyachamy, V. Alex, M. Khened, and G. Krishnamurthi, “Medical image retrieval using ResNet-18,” in Proc. SPIE Med. Imaging: Imaging Informatics Healthcare Res. Appl., San Diego, CA, USA, 2019, pp. 233–241.
20. [20] D. Sarwinda, R. Paradisa, A. Bustamam, and P. Anggia, “Deep learning in image classification using residual network (ResNet) variants for detection of colorectal cancer,” Procedia Comput. Sci., vol. 179, pp. 423–431, 2021.
21. [21] S. Xie, R. Girshick, P. Dollár, Z. Tu, and K. He, “Aggregated residual transformations for deep neural networks,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Honolulu, HI, USA, 2017, pp. 1492–1500.
22. [22] O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional networks for biomedical image segmentation,” in Proc. MICCAI, Munich, Germany, 2015, pp. 234–241.
23. [23] E. Zastrow, S. Davis, M. Lazebnik, F. Kelcz, B. Van Veen, and S. Hagness, “Database of 3D grid-based numerical breast phantoms for use in computational electromagnetics simulations,” Dept. Elect. Comput. Eng., Univ. Wisconsin-Madison, Madison, WI, USA, 2007.
24. [24] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Las Vegas, NV, USA, 2016, pp. 770–778.
25. [25] J. G. Proakis, Digital Signal Processing: Principles, Algorithms, and Applications, 4th ed., Upper Saddle River, NJ, USA: Pearson, 2007.