Modeling of Biomedical Antennas through Forecasting DNN for the Enlarged Bandwidth

• Published in: UK, Europe, China Millimetre Waves and THZ Technology Workshop (Online) pp. 223 - 226
• Main Authors: Kouhalvandi, Lida, Alibakhshikenari, Mohammad, Livreri, Patrizia, Matekovits, Ladislau, Peter, Ildiko
• Format: Conference Proceeding
• Language: English
• Published: IEEE 21.08.2024
• Subjects: Antennas; Bandwidth; biomedical; deep neural network (DNN); extended bandwidth; forecasting; implanted antenna; Long short term memory; long short-term memory (LSTM); multiple-input multiple-output (MIMO) antenna; Predictive models; Reflection coefficient; Reflector antennas; Scattering parameters; Terahertz materials; Topology; Wireless communication
• ISSN: 2639-4537
• DOI: 10.1109/UCMMT62975.2024.10737749

Recently, wireless medical technologies are growing day-by-day resulting in complex structures and topologies. Hence, advanced methods are required for designing and optimizing biomedical devices subject to high-dimensional parameter space. This paper is devoted to presenting an effective approach for estimating frequency responses of an implanted, multiple-input multiple-output (MIMO) antenna through the deep neural network (DNN) in terms of S11, S12, and total active reflection coefficient (TARC) specifications. This impressive approach aims to facilitate the time-consuming simulations in large multi-frequency bands and concurrently reduce the dependency on the designer's experience. All the process is performed in an automated environment and the proposed method is verified by designing and optimizing an implanted MIMO antenna operating in frequency bands of 4.34-4.61 GHz, and 5.86-6.64 GHz. In this design, the Long Short-Term Memory (LSTM)-based DNN is trained for the frequency band between 3-5.8 GHz, and afterward the constructed DNN is employed for predicting the various antenna specifications for the future bandwidth of 5.8-8 GHz.