Diffusional kurtosis imaging (DKI) has proven to be a promising diffusion-MRI technique whose first and most established applications are in neuroimaging. Recently, a number of preliminary studies have assessed the feasibility and potential usefulness of DKI in extra-cranial regions such as prostate, liver, kidney, bladder and breast. The stringent time constraints in most routine body MRI exams frequently mandate the acquisition of diffusion-weighted images (DWIs) with (only) three diffusion weighting directions (i.e. the main orthogonal directions). The aim of this study was to evaluate the potential error introduced in the estimation of the average of the three directional diffusional kurtosis values (K) by using, for each b-value, the geometric mean (trace-weighted image) of acquired DWIs (as is common practice in most diffusion-MRI studies of the body) instead of fitting the DKI model to DWIs acquired along each direction prior to averaging. By solving the DKI model analytically while imposing three orthogonal diffusion weighting directions and two non-null b-values (800 and 2000s/mm(2)), extensive simulations were performed for different K values (0-3) and a wide range of diffusion anisotropy values. The error in the estimates of K induced by geometrical averaging of DWIs was assessed and compared to the uncertainty in K caused by DWIs noise for low (20), medium (40) and high (80) signal-to-noise ratio (SNR) values. The simulations showed that geometrical averaging of the DWIs introduces a noticeable error in estimated K. While the error in K varies non-monotonically with K and with the degree of diffusion anisotropy, there is a trend of increasing absolute error with both increasing K and increasing degree of diffusion anisotropy. In particular, for values of K close to 1 and low/moderate (0-0.4) diffusion anisotropy degrees (typical of various body tissues), the absolute error in K can range up to 60% of K. In this case, at all SNR values (20, 40, 80), the absolute error in K can be greater than the uncertainty introduced by noise. In clinical body applications of DKI, the widespread and growing practice of using trace-weighted images to estimate K can introduce a substantial error, hence hampering interpretation of results as well as multi-center comparisons, and should therefore be avoided.
Giannelli, M., Toschi, N. (2016). On the use of trace-weighted images in body diffusional kurtosis imaging. MAGNETIC RESONANCE IMAGING, 34(4), 502-507 [10.1016/j.mri.2015.12.013].
On the use of trace-weighted images in body diffusional kurtosis imaging
TOSCHI, NICOLA
2016-01-01
Abstract
Diffusional kurtosis imaging (DKI) has proven to be a promising diffusion-MRI technique whose first and most established applications are in neuroimaging. Recently, a number of preliminary studies have assessed the feasibility and potential usefulness of DKI in extra-cranial regions such as prostate, liver, kidney, bladder and breast. The stringent time constraints in most routine body MRI exams frequently mandate the acquisition of diffusion-weighted images (DWIs) with (only) three diffusion weighting directions (i.e. the main orthogonal directions). The aim of this study was to evaluate the potential error introduced in the estimation of the average of the three directional diffusional kurtosis values (K) by using, for each b-value, the geometric mean (trace-weighted image) of acquired DWIs (as is common practice in most diffusion-MRI studies of the body) instead of fitting the DKI model to DWIs acquired along each direction prior to averaging. By solving the DKI model analytically while imposing three orthogonal diffusion weighting directions and two non-null b-values (800 and 2000s/mm(2)), extensive simulations were performed for different K values (0-3) and a wide range of diffusion anisotropy values. The error in the estimates of K induced by geometrical averaging of DWIs was assessed and compared to the uncertainty in K caused by DWIs noise for low (20), medium (40) and high (80) signal-to-noise ratio (SNR) values. The simulations showed that geometrical averaging of the DWIs introduces a noticeable error in estimated K. While the error in K varies non-monotonically with K and with the degree of diffusion anisotropy, there is a trend of increasing absolute error with both increasing K and increasing degree of diffusion anisotropy. In particular, for values of K close to 1 and low/moderate (0-0.4) diffusion anisotropy degrees (typical of various body tissues), the absolute error in K can range up to 60% of K. In this case, at all SNR values (20, 40, 80), the absolute error in K can be greater than the uncertainty introduced by noise. In clinical body applications of DKI, the widespread and growing practice of using trace-weighted images to estimate K can introduce a substantial error, hence hampering interpretation of results as well as multi-center comparisons, and should therefore be avoided.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
