Fast prediction of MRI RF-induced heating for a generic stent with arbitrary orientation using ANN

In this paper, a three-layer feed forward artificial neural network (ANN) is proposed to fast predict the radio frequency (RF)-induced heating for a generic stent with arbitrary posture under 1.5T MRI. The stent is studied of its RF-induced heating in various parameters, e.g. stent length, diameter and orientation, and the specific absorption rate (SAR) is adopted to characterize the heating level. 336 stent models are created and simulated based on the feasible stent parameters, 70% of which are utilized to train the ANN and the other 30% are utilized to test the performance of the ANN. It is shown that the simulated results and the predicted results are in good consistency. The maximum error between the target and the output is less than 7.6 W/kg. The correlation coefficient between the target and the output is larger than 0.998 for the testing data and 0.999 for all data.