一体化PET/MRI头部运动两种校正方法的对比

<正>电子发射断层成像(Position emission tomography,PET)/磁共振成像(Magnetic resonance imaging,MRI)问世以后,通过MRI获取扫描对象运动信息的准确度大为增加。为验证精确的运动信息条件下,基于响应线(Line of Response,LOR)的运动校正方法校正效果更为明显,我们对基于frame和基于LOR的两种头部运动校正方法的精度进行了对比研究。通过PET模拟数据完成两种校正算法的设计和结果比较,进而利用MRI成像速度快、图像质量高的特点,在上海联影医疗科技有限公司一体化PET/MRI上获取实验数据,通过配准MRI图像获取头部运动信息,并对PET数据进行运动补偿,实现对两种校正方法校正效果的验证和评估。通过对模拟和实验数据校正结果的定性定量分析,我们验证了PET/MRI环境中,基于LOR的方法,利用获取的高精度运动信息对于PET头部运动具有更好的补偿效果。

Evaluation of two motion correction methods for simultaneous PET/MRI brain imaging

Background:Simultaneous position emission tomography (PET) / magnetic resonance imaging (MRI) plays an important role in diagnosis of many brain diseases. However, brain motion caused by epilepsy, Parkinson's disease or muscle contraction and relaxation in head and neck is inevitable during the scanning. Motion artifact is one of the key factors that affect the quality of PET brain imaging. With PET/MRI, it becomes possible to use motion information obtained with MRI to correct for the PET image artifacts due to the high resolution of MRI.Purpose:This study aims to verify line of response (LOR) based motion correction method is more accurate than frame-based motion correction method in PET/MRI brain imaging, considering more precise information from MRI than previous methods.Methods: Rigid motions of NEMA (National Electrical Manufacturers Association) phantom and XCAT (the 4D extended cardiac-torso) phantom were simulated by using Monte Carlo method, i.e., the medical imaging simulation software GATE (GEANT4 application for tomographic emission). PET data was corrected using the two methods within open source reconstruction software STIR (software for tomographic image reconstruction). The reconstructed images of NEMA imaging quality (IQ) phantom were evaluated by contrast recovery coefficient (CRC) curves and the images of XCAT phantom were evaluated using full width at half maximum (FWHM) measurement of the lesion. The rigid phantom motion information was corrected by registering MRI images using gradient echo quick 3D sequence during PET/MRI scanning simultaneously, because MRI features high speed in imaging and high spatial resolution. Then, PET data was reconstructed using MRI derived motion vector to verify and evaluate the accuracy of these two motion correction methods.Results:FWHM values of reconstruction results compensated by both methods were significantly lower than the ones without motion correction. LOR based FWHM values were lower than those corrected by the frame-based methods both in XCAT simulation data and experiment data. Similarly, for NEMA IQ simulation data, the CRC curves had a higher upward tendency of both hot and cold spheres than the ones without motion correction, and the CRC curves of all spheres from LOR based method were higher than frame-based ones. Conclusion:By quantitative and qualitative analysis of both simulation and experiment corrected data, we concluded that both methods can compensate motion artifacts, and the LOR based method outperforms frame-based method for PET data compensation in simultaneous PET/MRI scanning.