Investigation of Artifacts and Optimization in Proton Resonance Frequency Thermometry Towards Heating Risk Monitoring of Implantable Medical Devices in Magnetic Resonance Imaging

• Published in: IEEE transactions on biomedical engineering 2021-12, Vol.68 (12), p.3638-3646
• Main Authors: Zhang, Feng, Jiang, Changqing, Li, Yichao, Niu, Xiaoyue, Long, Tiangang, He, Changgeng, Ding, Jianqi, Li, Linze, Li, Luming
• Format: Article
• Language: English
• Subjects: Bayesian analysis; Deep brain stimulation; Electrodes; Electronic implants; Heating; Heating systems; High temperature; Inhomogeneity; Magnetic fields; Magnetic properties; Magnetic resonance imaging; magnetic resonance thermometry; Medical equipment; Medical imaging; Neuroimaging; Normal distribution; Optimization; Performance enhancement; Phantoms; Proton resonance; proton resonance frequency; Radio frequency; radio frequency heating; Resonance; Temperature measurement; Thermometry
• ISSN: 0018-9294; 1558-2531
• DOI: 10.1109/TBME.2021.3081599

Objective : Artifacts limit the application of proton resonance frequency (PRF) thermometry for on-site, individualized heating evaluations of implantable medical devices such as deep brain stimulation (DBS) for use in magnetic resonance imaging (MRI). Its properties are unclear and the research on how to choose an unaffected measurement region is insufficient. Methods : The properties of PRF signals around the metallic DBS electrode were investigated through simulations and phantom experiments considering electromagnetic interferences from material susceptibility and the radio frequency (RF) interactions. A threshold method on phase difference Δ ϕ was used to define a measurement area to estimate heating at the electrode surface. Its performance was compared to that of the Bayesian magnitude method and probe measurements. Results : The B 0 magnetic field inhomogeneity due to the electrode susceptibility was the main influencing factor on PRF compared to the RF artifact. Δ ϕ around the electrode followed normal distribution but was distorted. Underestimation occurred at places with high temperature rises. The noise was increased and could be well estimated from magnitude images using a modified NEMA method. The Δ ϕ -threshold method based on this knowledge outperformed the Bayesian magnitude method by more than 42% in estimation error of the electrode heating. Conclusion : The findings favor the use of PRF with the proposed approach as a reliable method for electrode heating estimation. Significance : This study clarified the influence of device artifacts and could improve the performance of PRF thermometry for individualized heating assessments of patients with implants under MRI.