Magnetic resonance noise measurements and signal-quantization effects at very low noise levels

Mapping of magnetic resonance imaging's transverse relaxation time at low signal-to-noise ratio using Bloch simulations and principal component analysis image denoising

High-resolution mapping of magnetic resonance imaging (MRI)'s transverse relaxation time (T₂ ) can benefit many clinical applications by offering improved anatomic details, enhancing the ability to probe tissues' microarchitecture, and facilitating the identification of early pathology. Increasing spatial resolutions, however, decreases data's signal-to-noise ratio (SNR), particularly at clinical scan times. This impairs imaging quality, and the accuracy of subsequent radiological interpretation. Recently, principal component analysis (PCA) was employed for denoising diffusion-weighted MR images and was shown to be effective for improving parameter estimation in multiexponential relaxometry. This study combines the Marchenko-Pastur PCA (MP-PCA) signal model with the echo modulation curve (EMC) algorithm for denoising multiecho spin-echo (MESE) MRI data and improving the precision of EMC-generated single T₂ relaxation maps. The denoising technique was validated on simulations, phantom scans, and in vivo brain and knee data. MESE scans were performed on a 3-T Siemens scanner. The acquired images were denoised using the MP-PCA algorithm and were then provided as input for the EMC T₂ -fitting algorithm. Quantitative analysis of the denoising quality included comparing the standard deviation and coefficient of variation of T₂ values, along with gold standard SNR estimation of the phantom scans. The presented denoising technique shows an increase in T₂ maps' precision and SNR, while successfully preserving the morphological features of the tissue. Employing MP-PCA denoising as a preprocessing step decreases the noise-related variability of T₂ maps produced by the EMC algorithm and thus increases their precision. The proposed method can be useful for a wide range of clinical applications by facilitating earlier detection of pathologies and improving the accuracy of patients' follow-up.

Higher b-values improve the correlation between diffusion MRI and the cortical microarchitecture

PURPOSE

In diffusion MRI (dMRI), it remains unclear to know how much increase of b-value is conveying additional biological meaning. We tested the correlations between cortical microarchitecture and diffusion metrics computed from standard (1000 s/mm²), high (3000 s/mm²), to very high (5000 s/mm²) b-value dMRI.

METHODS

Healthy volunteers were scanned with a dMRI pulse sequence that was first optimized together with a T1-WI and T2-WI. Averaged cortical surface map of estimated myelin (T1-WI/T2-WI) was compared with surface maps of mean diffusivity (MD) computed from each b-value (MD₁₀₀₀, MD₃₀₀₀, and MD₅₀₀₀) and to surface map of mean kurtosis (MK computed from the 0-, 1000-, to 3000-s/mm² shells) in 360 cortical parcels using Spearman correlations, multiple linear regressions, and Akaike information criteria (AIC).

RESULTS

Surface map from MD₁₀₀₀ showed variations not related to myelin but the MD₃₀₀₀ and MD₅₀₀₀ maps inversely mirrored estimated myelin map; lower MD values being observed in more myelinated cortical areas. MK mirrored myelinated cortical areas. Quantitatively, Spearman correlations between myelin and MD became more and more negative as long as b-values increased while the correlation was positive between myelin and MK. Multiple regression models confirmed negative associations between myelin and MD that were significantly better from MD₁₀₀₀ to MD₃₀₀₀ and MD₅₀₀₀ (R² = 0.33, p < 0.001; R² = 0.43, p < 0.001; and R² = 0.50, p < 0.001) and positive association between myelin and MK (R² = 0.53, p < 0.001). Comparisons of the 3 statistical models showed the best performances with MK and MD₅₀₀₀ (AIC_MK < AIC_MD5000 < AIC_MD3000 < AIC_MD1000).

CONCLUSION

Higher b-values are more closely related to subtle cellular variations of the cortical microarchitecture.

Impact of contrast agent injection duration on dynamic contrast-enhanced MRI quantification in prostate cancer

The volume transfer constant K^trans , which describes the leakage of contrast agent (CA) from vasculature into tissue, is the most commonly reported quantitative parameter for dynamic contrast-enhanced (DCE-) MRI. However, the variation in reported K^trans values between studies from different institutes is large. One of the primary sources of uncertainty is quantification of the arterial input function (AIF). The aim of this study is to determine the influence of the CA injection duration on the AIF and tracer kinetic analysis (TKA) parameters (i.e. K^trans , k_ep and v_e ). Thirty-one patients with prostate cancer received two DCE-MRI examinations with an injection duration of 5 s in the first examination and a prolonged injection duration in the second examination, varying between 7.5 s and 30 s. The DCE examination was carried out on a 3.0 T MRI scanner using a transversal T₁ -weighted 3D spoiled gradient echo sequence (300 s duration, dynamic scan time of 2.5 s). Data of 29 of the 31 were further analysed. AIFs were determined from the phase signal in the left and right femoral arteries. K^trans , k_ep and v_e were estimated with the standard Tofts model for regions of healthy peripheral zone and tumour tissue. We observed a significantly smaller peak height and increased width in the AIF for injection durations of 15 s and longer. However, we did not find significant differences in K^trans , k_ep or v_e for the studied injection durations. The study demonstrates that the TKA parameters K^trans , k_ep and v_e , measured in the prostate, do not show a significant change as a function of injection duration.

Internal evaluation of impregnation treatment of waterlogged wood; relation between concentration of internal materials and relaxation time using magnetic resonance imaging

The purpose of this study is to clarify the degree of impregnation resulting from treatment of internal waterlogged wood samples using MRI. On a 1.5T MR scanner, T₁ and T₂ measurements were performed using inversion recovery and spin-echo sequences, respectively. The samples were cut waterlogged pieces of wood treated with various impregnation techniques which were divided into different concentrations of trehalose (C₁₂H₂₂O₁₁) and polyethylene glycol (PEG; HO-(C₂H₄O)_n-H) solutions. Then these samples underwent impregnation treatment every two weeks. From the results, we found that the slope of the T₁-concentration curve using linear fitting showed the value of the internal area for PEG to be higher than the external area; internal, -2.73ms/wt% (R²=0.880); external, -1.50ms/wt% (R²=0.887). Furthermore, the slope of the T₁-concentration curve using linear fitting showed the values for trehalose to have almost no difference when comparing the internal and the external areas; internal, -2.79ms/wt% (R²=0.759); external, -3.02ms/wt% (R²=0.795). However, the slope of the T₂-concentration curve using linear fitting for PEG showed that there was only a slight change between the internal and the external areas; internal, 0.26ms/wt% (R²=0.642); external, 0.18ms/wt% (R²=0.920). The slope of the T₂-concentration curve did not show a change in linear relationship between the internal and the external areas; internal, 0.06ms/wt% (R²=0.175); external, -0.14ms/wt% (R²=0.043). In conclusion, using visualization of relaxation time T₁, it is possible to obtain more detail information noninvasively concerning the state of impregnation treatment of internal waterlogged wood.

Fast volumetric imaging of bound and pore water in cortical bone using three-dimensional ultrashort-TE (UTE) and inversion recovery UTE sequences

We report the three-dimensional ultrashort-TE (3D UTE) and adiabatic inversion recovery UTE (IR-UTE) sequences employing a radial trajectory with conical view ordering for bi-component T2 * analysis of bound water (T2 *(BW) ) and pore water (T2 *(PW) ) in cortical bone. An interleaved dual-echo 3D UTE acquisition scheme was developed for fast bi-component analysis of bound and pore water in cortical bone. A 3D IR-UTE acquisition scheme employing multiple spokes per IR was developed for bound water imaging. Two-dimensional UTE (2D UTE) and IR-UTE sequences were employed for comparison. The sequences were applied to bovine bone samples (n = 6) and volunteers (n = 6) using a 3-T scanner. Bi-component fitting of 3D UTE images of bovine samples showed a mean T2 *(BW) of 0.26 ± 0.04 ms and T2 *(PW) of 4.16 ± 0.35 ms, with fractions of 21.5 ± 3.6% and 78.5 ± 3.6%, respectively. The 3D IR-UTE signal showed a single-component decay with a mean T2 *(BW) of 0.29 ± 0.05 ms, suggesting selective imaging of bound water. Similar results were achieved with the 2D UTE and IR-UTE sequences. Bi-component fitting of 3D UTE images of the tibial midshafts of healthy volunteers showed a mean T2 *(BW) of 0.32 ± 0.08 ms and T2 *(PW) of 5.78 ± 1.24 ms, with fractions of 34.2 ± 7.4% and 65.8 ± 7.4%, respectively. Single-component fitting of 3D IR-UTE images showed a mean T2 *(BW) of 0.35 ± 0.09 ms. The 3D UTE and 3D IR-UTE techniques allow fast volumetric mapping of bound and pore water in cortical bone. Copyright © 2016 John Wiley & Sons, Ltd.

A simple noise correction scheme for diffusional kurtosis imaging

PURPOSE

Diffusional kurtosis imaging (DKI) is sensitive to the effects of signal noise due to strong diffusion weightings and higher order modeling of the diffusion weighted signal. A simple noise correction scheme is proposed to remove the majority of the noise bias in the estimated diffusional kurtosis.

METHODS

Weighted linear least squares (WLLS) fitting together with a voxel-wise, subtraction-based noise correction from multiple, independent acquisitions are employed to reduce noise bias in DKI data. The method is validated in phantom experiments and demonstrated for in vivo human brain for DKI-derived parameter estimates.

RESULTS

As long as the signal-to-noise ratio (SNR) for the most heavily diffusion weighted images is greater than 2.1, errors in phantom diffusional kurtosis estimates are found to be less than 5 percent with noise correction, but as high as 44 percent for uncorrected estimates. In human brain, noise correction is also shown to improve diffusional kurtosis estimates derived from measurements made with low SNR.

CONCLUSION

The proposed correction technique removes the majority of noise bias from diffusional kurtosis estimates in noisy phantom data and is applicable to DKI of human brain. Features of the method include computational simplicity and ease of integration into standard WLLS DKI post-processing algorithms.

Incorporation of Rician noise in the analysis of biexponential transverse relaxation in cartilage using a multiple gradient echo sequence at 3 and 7 Tesla

PURPOSE

Previous work has evaluated the quality of different analytic methods for extracting relaxation times from magnitude imaging data exhibiting Rician noise. However, biexponential analysis of relaxation in tissue, including cartilage, and materials is of increasing interest. We, therefore, analyzed biexponential transverse relaxation decay in the presence of Rician noise and assessed the accuracy and precision of several approaches to determining component fractions and apparent transverse relaxation times.

THEORY AND METHODS

Comparisons of four different voxel-by-voxel fitting methods were performed using Monte Carlo simulations, and phantom and ex vivo bovine nasal cartilage (BNC) experiments. In each case, preclinical and clinical imaging field strengths of 7 Tesla (T) and 3T, respectively, and parameters, were investigated across a range of signal-to-noise ratios (SNR). Results were compared with Cramér-Rao lower bound calculations.

RESULTS

As expected, at high SNR, all methods performed well. At lower SNR, fits explicitly incorporating the analytic form of the Rician noise maintained performance. The much more efficient correction scheme of Gudbjartsson and Patz performed almost as well in many cases. Ex vivo experiments on phantoms and BNC were consistent with simulation results.

CONCLUSION

Explicit incorporation of Rician noise greatly improves accuracy and precision in the analysis of biexponential transverse decay data.

Improved R2* measurement accuracy with absolute SNR truncation and optimal coil combination

Accurate R2* measurements are critical for many abdominal imaging applications. Conventionally, R2* maps are derived via the monoexponential fitting of signal decay within a series of gradient-echo (GRE) images reconstructed from multichannel datasets combined using a root sum-of-squares (RSS) approach. However, the noise bias at low-SNR TEs from RSS-reconstructed data often causes the underestimation of R2* values. In phantom, ex vivo animal model and normal volunteer studies, we investigated the accuracy of low-SNR R2* measurement when combining truncation and coil combination methods. The accuracy for R2* estimations was shown to be affected by the intrinsic R2* value, SNR level and the chosen reconstruction method. The R2* estimation error was found to decrease with increasing SNR level, decreasing R2* value and the use of the optimal B1-weighted combined (OBC) image reconstruction method. Data truncation based on rigorous voxel-wise SNR estimates can reduce R2* measurement error in the setting of low SNR with fast signal decay. When optimal SNR truncation thresholds are unknown, the OBC method can provide optimal R2* measurements given the minimal truncation requirements.