Cochlear implant artifact analysis in cranial MRI: effects of field strength and imaging sequence.
Authors
Affiliations (8)
Affiliations (8)
- Department of Otorhinolaryngology, Affiliated Hospital of Hangzhou Normal University, Hangzhou, Zhejiang, China.
- Hangzhou Normal University, Hangzhou, Zhejiang, China.
- Department of Radiology, Affiliated Hospital of Hangzhou Normal University, Hangzhou, Zhejiang, China.
- College of Biomedical Engineering and Instrument Science, Zhejiang University, Hangzhou, Zhejiang, China.
- Zhejiang Nurotron Biotechnology Co., Ltd., Hangzhou, Zhejiang, China.
- Central Hospital, Jinhua, Zhejiang, China.
- School of Physics, Zhejiang University, Hangzhou, Zhejiang, China.
- Yiwu The Center for Caribbean Studies at Hangzhou Normal University, Hangzhou, Zhejiang, China.
Abstract
To investigate the artifacts produced by the cochlear implant (CI) CS-30A in nonimplanted human clinical trials under different magnetic resonance (MR) field strengths and imaging sequences. The CI was placed at a standardized surface projection location on the temporal region of 12 healthy subjects. Axial cranial MR scans were performed at both 1.5 T and 3.0 T field strengths using the following sequences: T1-weighted spin-echo (SE), T2-weighted fast spin-echo (FSE), and T2*-weighted gradient-echo (GRE). The maximum dimension of CI artifacts was measured, and the artifact area was quantified using both manual measurement and a deep learning-based automatic segmentation algorithm. The impact of CI artifacts on the visibility of brain anatomy was evaluated by radiologists. Under both 1.5 T and 3.0 T MR scanning, GRE sequences exhibited significantly larger artifact maximum dimension and area than SE and FSE sequences (<i>p</i> < 0.05). No significant differences were noted in maximum artifact dimensions between 1.5 T and 3.0 T MR across sequences (<i>p</i> > 0.05). The difference in artifact area between manual measurement and the deep learning-based automatic segmentation algorithm was not significant on SE sequences (<i>p</i> = 0.377). GRE sequences produced a wider range of CI artifacts than SE and FSE sequences at both 1.5 T and 3.0 T. No significant difference in artifact size was observed between the two field strengths. The artifact areas on SE sequences showed high consistency between manual measurement and the deep learning-based automatic segmentation algorithm. Therefore, SE or FSE sequences should be prioritized over GRE for CI recipients, and both 1.5 T and 3.0 T are clinically feasible.