Noise Removal in Magnetic Resonance Diffusion Tensor Imaging英文文献.pdfVIP

Magnetic Resonance in Medicine 54:393–407 (2005) Noise Removal in Magnetic Resonance Diffusion Tensor Imaging Bin Chen1 and Edward W. Hsu1,2* Although promising for visualizing the structure of ordered tis- porating the diffusion-independent signal intensity via an sues, MR diffusion tensor imaging (DTI) has been hampered by additional variable, a unique solution to the multivariate long acquisition time and low spatial resolution associated with system thus requires a minimum of seven image acquisi- its inherently low signal-to-noise ratio (SNR). Moreover, the tions. uncertainty in the DTI measurements has a direct impact on the Due to the signal-attenuation nature of diffusion encod- accuracy of structural renderings such as fiber streamline ing and the inadvertent T2 relaxation resulting from the tracking. Noise removal techniques can be used to improve the SNR of DTI without requiring additional acquisitions, albeit long TE necessary to accommodate the sensitization gra- most low-pass filtering methods are accompanied by undesir- dient pulses, the signal-to-noise ratio (SNR) of DTI is in- able image blurring. In the present study, a modified vector- herently low. Moreover, because of the large data set size based partial-differential-equation (PDE) filtering formalism and the tradeoff between SNR and resolution, applications was implemented for smoothing DTI vector fields. Using an of DTI have been challenged by poor SNR and low spatial image residual-energy criterion to equate the degree of and temporal resolutions. Adverse impacts on DTI include smoothing and error metrics empiric