Neural networks modeling of temperature field distribution in hyperthermia

Abstract

Hyperthermia is known to be a method of killing tumor cells by heating. An ultrasound transducer is often used as the heating device. In order to kill the tumor cells and not injure the normal tissue, the temperature distribution generated by the ultrasound must be predetermined. For a multi-element ultrasound transducer, the phase and the amplitude of the input signal for each element can be tuned to generate a suitable temperature distribution to meet the needs of individual treatments. However, direct computation is often time-consuming, while there are also difficulties in computing the ultrasound transducer parameters with a given temperature distribution. In this paper, artificial neural networks are used to learn the relationship between the ultrasound transducer parameters and the temperature distribution, both in the forward and in the inverse direction.