T(2) decay curve acquisition and analysis in MRI: Noise considerations, short T(2), and B(1) field encoding.

摘要

The myelin sheath is a lipid bilayer wrapped around axons in the human brain. The myelin allows faster conduction and uses less energy than along non-myelinated axons. Diseases, such as multiple sclerosis, break down the myelin sheath resulting in cognitive and physical disability. Water trapped between the myelin bilayers has a T₂ ≈ 15 ms compared to intra/extra-cellular water with a T₂ ≈ 80 ms and cerebrospinal fluid which has a T₂ ≈ 2000 ms. A multi-echo MRI pulse sequence is used to acquire a T₂ decay curve which is fit using a non-negative least-squares (NNLS) curve fitting algorithm to calculate the signal as a function of the T₂ (T₂ distribution). The myelin water fraction (MWF) is calculated as the signal with a T₂ 50 ms divided by the total signal in the T₂ distribution. Quantification of the MWF in vivo, is slow, prone to errors due to artifacts in the T ₂ decay curve, and requires many data acquisition averages to attain the necessary high signal-to-noise ratio. Each of these areas were addressed in this thesis. First, I compared the accuracy and precision of MWF estimates from a set of simulated and in vivo multi-echo data with and without noise reduction filters applied. The MWF estimated from anisotropically filtered multi-echo data had the smallest variability; over homogeneous regions, compared to MWF estimates from other filtered data. Second, I created a new technique to optimize coefficients that when linearly combined with multi-echo data, result in fast, accurate estimates of the MWF. Simulations and in vivo estimates of the MWF were as accurate as those from the NNLS algorithm and 20,000 times faster. Finally, the standard multi-echo sequence was optimized to remove artifacts due to inaccurate refocusing pulses. I created a novel T₂ curve fitting algorithm, based on NNLS, to accurately estimate the T₂ distribution and refocusing pulse flip angle from a T₂ decay curve. Simulations and data acquired on phantoms showed the new technique was as accurate as the NNLS algorithm in quantifying the T₂ distribution parameters. In vivo myelin water maps were as good as those estimated from the optimized multi-echo pulse sequence.