Geometric distortion in magnetic resonance imaging systems assessed using an open-source plugin for scientific image analysis.
Takahiro AoyamaHidetoshi ShimizuIkuo ShimizuAtsushi TeramotoNaoki KanedaKazuhiko NakamuraMasaru NakamuraTakeshi KodairaPublished in: Radiological physics and technology (2018)
Tumor locations are commonly delineated by referring to magnetic resonance (MR) images. However, MR images have geometric distortions that cannot be completely corrected. This study aimed to investigate quantitatively uncorrectable error [residual error (RE)] with the use of an open-source plugin for scientific image analysis. The RE values were calculated by Fiji, which was enhanced by Image J image processing software. The results obtained with the open-source plugin for scientific image analysis agreed with the results obtained with the commercially available software. Obtaining detailed geometric distortion data for each facility and device could facilitate safe treatment because the homogeneous magnetic field in MR imaging varies across devices and over time. Therefore, using an open-source plugin for scientific image analysis may be an accurate and effective technique for evaluating the RE of MR imaging systems.
Keyphrases
- contrast enhanced
- magnetic resonance
- magnetic resonance imaging
- deep learning
- convolutional neural network
- computed tomography
- diffusion weighted imaging
- optical coherence tomography
- artificial intelligence
- data analysis
- electronic health record
- machine learning
- combination therapy
- smoking cessation
- replacement therapy