Application of Machine learning to predict RF heating of cardiac leads during magnetic resonance imaging at 1.5 T and 3 T: A simulation study

Abstract

Predicting magnetic resonance imaging (MRI)-induced heating of elongated conductive implants, such as leads in cardiovascular implantable electronic devices, is essential to assessing patient safety. Phantom experiments have traditionally been used to estimate radio-frequency (RF) heating of implants, but they are time-consuming. Recently, machine learning has shown promise for fast prediction of RF heating of orthopaedic implants when the implant position within the MRI RF coil was predetermined. We explored whether deep learning could be applied to predict RF heating of conductive leads with variable positions and orientations during MRI at 1.5 T and 3 T. Models of 600 cardiac leads with clinically relevant trajectories were generated, and electromagnetic simulations were performed to calculate the maximum of the 1 g-averaged specific absorption rate (SAR) of RF energy at the tips of lead models during MRI at 1.5 T and 3 T. Neural networks were trained to predict the maximum SAR at the lead tip from the knowledge of the coordinates of points along the lead trajectory. Despite the large range of SAR values (∼230 W/kg to ∼ 3200 W/kg and ∼ 10 W/kg to ∼ 3300 W/kg), the root- mean-square error of the predicted vs ground truth SAR remained at 223 W/kg and 206 W/kg, with the R2 scores of 0.89 and 0.85 on the testing set for 1.5 T and 3 T models, respectively. The results suggest that machine learning is a promising approach for fast assessment of RF heating of lead-like implants when only the knowledge of the lead geometry and MRI RF coil features are in hand.