Quantification Of Transverse Relaxation In Magnetic Resonance Imaging

Magnetic resonance imaging?MRI?is a novel medical imaging technique developed in 1980s.After decades of development,MRI plays an important role in medical imaging diagnosis and scientific research,and is widely applied to medical diagnosis,staging of disease and prognosis evaluation.Iron is an essential trace element in the human body,however,excessive iron accumulation promotes the formation of toxic oxygen radicals that can lead to cell demage.For patients with thalassemia and sickle cell disease,chronic blood transfusions can also develop iron overload.Though excessive iron preferentially accumulates within Kupffer cells rather than hepatocytes,the damaging effects occur in heart and endocrine system,which may lead to cardiomyopathy and cardiac arrhythmias due to cardiac toxicity.For those iron overloaded livers,therapeutic phlebotomy or treatment with chelators are usually performed to remove excess iron.Thus,accurate measurement of liver iron overload is of clinical significance for instructing and montoring the treatment.Considering that excessive iron is mainly deposited in liver and liver iron concentration?LIC?is commonly used as a surrogate for the total body iron loading,thus biopsy or MRI is usually performed on liver.Percutaneous biopsy is the current gold standard for the evaluation of LIC,however it is an invasive,painful,and expensive procedure with 1%to 4%risk of potentially serious complications.MRI based R₂ or R₂*techniques have emerged as reliable alternatives,and the relationships between R₂ or R₂*and biopsied LIC have been well described.The remaining challenges include 1)in scenarios of high LIC or low signal-to-noise-ratio?SNR?,current signal model is not accurate enough and may lead to biased R₂ or R₂*estimations.2)relaxivity-iron calibrations are mainly performed at 1.5T?Tesla?with specific imaging sequences,and results at higher magnetic fields or with other sequences are still unknown.3)fat in the liver can also affect R₂*quantification,though the impact of fat on R₂*is small when compared with iron overload,and the best possible accuracy is important for patients with mild-moderate iron overload and high fat content.To address the above issues,this dissertation focuses on quantification of transverse relaxations in MRI.Especially,we investigate the accurate R₂*mapping under different noise levels and varying iron overload levels,as well as relaxivity-iron calibration and relaxivity-fat calibration.The research contents in this dissertation include the following four aspects:?1?Rapid look-up table method has been proposed for noise-corrected curve fitting in the R₂*mapping of iron loaded liver.For liver data acquired by multichannel array coils and reconstructed by root-sum-square operation,the noise is generally assumed to follow a noncentral chi distribution.Fitting the measured decay signal to the first moment in the presence of noise?M¹NCM?is shown to be capable of correctly addressing the effect of the noncentral chi noise on R₂*relaxometry.However,fitting by the M¹NCM model requires intensive computation of the confluent hypergeometric function,which limits its application to pixelwise fitting based R₂*mapping.To accelerate R₂*mapping with the M¹NCM model,the proposed method approximates the confluent hypergeometric function using the cubic spline interpolant with breakpoints and coefficients precalculated and stored in a look-up table,the performance of the proposed method is evaluated on both simulation and in vivo data of the liver R₂*relaxometry.Results demonstrate that the proposed method can calculate one R₂*map in 1 s and is approximately five orders of magnitude faster than the standard M¹NCM method while generating nearly identical R₂*maps.?2?Pixel-wise curve fitting with adaptive neighborhood regularization has been proposed to improve liver R₂*mapping.Magnetic resonance imaging R₂*mapping remains challenging because of the serial images with low SNR levels.We propose to exploit the neighboring pixels as regularization terms and adaptively determine the regularization parameters according to the interpixel signal similarity.The proposed algorithm is evaluated on simulated,phantom and in vivo data.Results demonstrate that the proposed algorithm generates R₂*maps with significantly reduced noise and well-preserved tiny structures.Quantitatively,the proposed algorithm can produce R₂*maps with lower root mean square errors,and high accuracy and precision in terms of mean and standard deviation of R₂*measurements in selected region of interests,at varying R₂*values and SNR levels.?3?A Monte Carlo model has been reproduced and extended for relaxivity-iron calibration in hepatic iron overload.Relaxation rates R₂*and single spin echo R₂ at 1.5T have been empirically calibrated to biopsy-measured LIC,however,the underlying bio-physical mechanism is still unknown.In addition,calibrations at higher magnetic fields are unavailable.Relaxivity-iron calibration curves at 1.5T and 3.0T are simulated using a previously developed Monte Carlo model.Furthermore,the model is extended with multiple spin echo imaging,and iron calibrations are evaluated using two different fitting models.Results demonstrate that relaxivity-iron calibration is reproducible using the developed Monte Carlo model,and the model can be readily extended to other important applications,including predicting signal behavior for multiple spin echo imaging.?4?Liver MR signal in the presence of fat has been investigated using Monte Carlo modeling.Chemical shift-encoded MRI based proton density fat fraction?PDFF?and R₂*quantification in the liver are promising biomarkers for triglyceride concentration and iron concentration,respectively.In previous studies,a positive correlation has been observed between PDFF and R₂*in the absence of iron overload.Phantom experiments and Monte Carlo simulation with different fat susceptibilities are performed to determine the underlying mechanism of the impact of fat on R₂*at 1.5T and 3.0T.Results demonstrate that the clinically observed correlation between PDFF and R₂*is successfully recapitulated using the Monte Carlo simulations with realistic physical parameters.