Bootstrap-combined Data Analysis And Application Of Magnetic Resonance Spectroscopy And T1 Mapping

Magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS) are non-invasive and non-radiant techniques in biomedical imaging domain. The resolution of MRI and MRS is also better than ultrasound imaging (UI), positron emission tomography (PET) and other imaging modalities. What is more, in 1979 the American scientist Efron systematically introduced the Bootstrap technique as a method to deduce the standard error of random estimator. Since then, the Bootstrap method has been used in a wide range of subject areas such as biomedicine, economics, medicine and foreign trade.The MRS research was based on Residual Bootstrap. The innovation point of this part is the application of Residual Bootstrap to estimate the uncertainty of GABA concentration and obtain a robust result.Spectral editing MEGA-PRESS sequence and GABA Analysis Toolkit Gannet were used to detect and analyze GABA concentration in anterior cingulate cortex (ACC) and occipital cortex (OCC) on 12 healthy volunteers. They were all scanned twice with a time interval of at least 1 week. Residual Bootstrap, a model-based technique, was applied to resample fitting residual of GABA from the Gaussian fitting model in both individual and group data of ACC and OCC.With the usage of Residual Bootstrap, the coefficient of variation (CV) of GAB A+in OCC was lower than in ACC in the intra-subject study, while higher than in ACC in the inter-subject study. What’s more, the CV value and uncertainty after Residual Bootstrap were all lower than without the application of Residual Bootstrap method.The Residual Bootstrap analysis provides a robust uncertainty estimation of individual and group GABA detection in different brain regions.The T1 mapping research was based on Wild Bootstrap. The innovation points of this part is the comparison of uncertainty estimation of T1 values by MR scans in different sites with the Wild Bootstrap method.In the Hangzhou First People’s Hospital (site 1) and Sir Run Run Shaw Hospital(site 2), a 3.0 T Siemens and a 3.0 T GE MR scanners were used respectively. Polyvinylpyrrolidone (PVP) solutions with different concentration and the brain of a healthy male volunteer were been scanned several times to acquire the T1 mapping with the DESPOT method. Wild Bootstrap method was used to estimate the uncertainty of intra-, inter-site and within-scan.Both the PVP and human brain data showed that, the shorter T1 values, the smaller uncertainty and CV values. Comparing with the uncertainty of site 2, data was found more stable in site 1.The Wild Bootstrap analysis is necessary and helpful for multi-center T1 quantification studies, and the intra and inter site uncertainty calculation also provide a robust evaluation for quality assurance and T1 mapping protocols optimization.