Reliable estimation of brain intravoxel incoherent motion parameters using denoised diffusion‐weighted MRI

In this study, we evaluate whether diffusion‐weighted magnetic resonance imaging (DW‐MRI) data after denoising can provide a reliable estimation of brain intravoxel incoherent motion (IVIM) perfusion parameters. Brain DW‐MRI was performed in five healthy volunteers on a 3 T clinical scanner with 12 different b‐values ranging from 0 to 1000 s/mm². DW‐MRI data denoised using the proposed method were fitted with a biexponential model to extract perfusion fraction (PF), diffusion coefficient (D) and pseudo‐diffusion coefficient (D*). To further evaluate the accuracy and precision of parameter estimation, IVIM parametric images obtained from one volunteer were used to resimulate the DW‐MRI data using the biexponential model with the same b‐values. Rician noise was added to generate DW‐MRI data with various signal‐to‐noise ratio (SNR) levels. The experimental results showed that the denoised DW‐MRI data yielded precise estimates for all IVIM parameters. We also found that IVIM parameters were significantly different between gray matter and white matter (P < 0.05), except for D* (P = 0.6). Our simulation results show that the proposed image denoising method displays good performance in estimating IVIM parameters (both bias and coefficient of variation were <12% for PF, D and D*) in the presence of different levels of simulated Rician noise (SNR_b=0 = 20‐40). Simulations and experiments show that brain DW‐MRI data after denoising can provide a reliable estimation of IVIM parameters.     

Stroke assessment with intravoxel incoherent motion diffusion‐weighted MRI

Intravoxel incoherent motion (IVIM) diffusion‐weighted MRI can simultaneously measure diffusion and perfusion characteristics in a non‐invasive way. This study aimed to determine the potential utility of IVIM in characterizing brain diffusion and perfusion properties for clinical stroke. The multi‐b‐value diffusion‐weighted images of 101 patients diagnosed with acute/subacute ischemic stroke were retrospectively evaluated. The diffusion coefficient D, representing the water apparent diffusivity, was obtained by fitting the diffusion data with increasing high b‐values to a simple mono‐exponential model. The IVIM‐derived perfusion parameters, pseudodiffusion coefficient D*, vascular volume fraction f and blood flow‐related parameter fD*, were calculated with the bi‐exponential model. Additionally, the apparent diffusion coefficient (ADC) was fitted according to the mono‐exponential model using all b‐values. The diffusion parameters for the ischemic lesion and normal contralateral region were measured in each patient. Statistical analysis was performed using the paired Student t‐test and Pearson correlation test. Diffusion data in both the ischemic lesion and normal contralateral region followed the IVIM bi‐exponential behavior, and the IVIM model showed better goodness of fit than the mono‐exponential model with lower Akaike information criterion values. The paired Student t‐test revealed significant differences for all diffusion parameters (all P < 0.001) except D* (P = 0.218) between ischemic and normal areas. For all patients in both ischemic and normal regions, ADC was significantly positively correlated with D (both r = 1, both P < 0.001) and f (r = 0.541, P < 0.001; r = 0.262, P = 0.008); significant correlation was also found between ADC and fD* in the ischemic region (r = 0.254, P = 0.010). For all pixels within the region of interest from a representative subject in both ischemic and normal regions, ADC was significantly positively correlated with D (both r = 1, both P < 0.001), f (r = 0.823, P < 0.001; r = 0.652, P < 0.001) and fD* (r = 0.294, P < 0.001; r = 0.340, P < 0.001). These findings may have clinical implications for the use of IVIM imaging in the assessment and management of acute/subacute stroke patients. Copyright © 2016 John Wiley & Sons, Ltd.     

Relationship between diffusion parameters derived from intravoxel incoherent motion MRI and perfusion measured by dynamic contrast‐enhanced MRI of soft tissue tumors

Our aim was to evaluate the link between diffusion parameters measured by intravoxel incoherent motion (IVIM) diffusion‐weighted imaging (DWI) and the perfusion metrics obtained with dynamic contrast‐enhanced (DCE) MRI in soft tissue tumors (STTs). Twenty‐eight patients affected by histopathologically confirmed STT were included in a prospective study. All patients underwent both DCE MRI and IVIM DWI. The perfusion fraction f, diffusion coefficient D and perfusion‐related diffusion coefficient D* were estimated using a bi‐exponential function to fit the DWI data. DCE MRI was acquired with a temporal resolution of 3–5 s. Maps of the initial area under the gadolinium concentration curve (IAUGC), time to peak (TTP) and maximum slope of increase (MSI) were derived using commercial software. The relationships between the DCE MRI and IVIM DWI measurements were assessed by Spearman's test. To exclude false positive results under multiple testing, the false discovery rate (FDR) procedure was applied. The Mann–Whitney test was used to evaluate the differences between all variables in patients with non‐myxoid and myxoid STT. No significant relationship was found between IVIM parameters and any DCE MRI parameters. Higher f and D*f values were found in non‐myxoid tumors compared with myxoid tumors (p = 0.004 and p = 0.003, respectively). MSI was significantly higher in non‐myxoid tumors than in myxoid tumors (p = 0.029). From the visual assessments of single clinical cases, both f and D*f maps were in satisfactory agreement with DCE maps in the extreme cases of an avascular mass and a highly vascularized mass, whereas, for tumors with slight vascularity or with a highly heterogeneous perfusion pattern, this association was not straightforward. Although IVIM DWI was demonstrated to be feasible in STT, our data did not support evident relationships between perfusion‐related IVIM parameters and perfusion measured by DCE MRI. Copyright © 2015 John Wiley & Sons, Ltd.     

Triggered intravoxel incoherent motion MRI for the assessment of calf muscle perfusion during isometric intermittent exercise

The main aim of this paper was to propose triggered intravoxel incoherent motion (IVIM) imaging sequences for the evaluation of perfusion changes in calf muscles before, during and after isometric intermittent exercise. Twelve healthy volunteers were involved in the study. The subjects were asked to perform intermittent isometric plantar flexions inside the MRI bore. MRI of the calf muscles was performed on a 3.0 T scanner and diffusion‐weighted (DW) images were obtained using eight different b values (0 to 500 s/mm²). Acquisitions were performed at rest, during exercise and in the subsequent recovery phase. A motion‐triggered echo‐planar imaging DW sequence was implemented to avoid movement artifacts. Image quality was evaluated using the average edge strength (AES) as a quantitative metric to assess the motion artifact effect. IVIM parameters (diffusion D, perfusion fraction f and pseudo‐diffusion D*) were estimated using a segmented fitting approach and evaluated in gastrocnemius and soleus muscles. No differences were observed in quality of IVIM images between resting state and triggered exercise, whereas the non‐triggered images acquired during exercise had a significantly lower value of AES (reduction of more than 20%). The isometric intermittent plantar‐flexion exercise induced an increase of all IVIM parameters (D by 10%; f by 90%; D* by 124%; fD* by 260%), in agreement with the increased muscle perfusion occurring during exercise. Finally, IVIM parameters reverted to the resting values within 3 min during the recovery phase. In conclusion, the IVIM approach, if properly adapted using motion‐triggered sequences, seems to be a promising method to investigate muscle perfusion during isometric exercise.     

Quantification of microcirculatory parameters by joint analysis of flow‐compensated and non‐flow‐compensated intravoxel incoherent motion (IVIM) data

The aim of this study was to improve the accuracy and precision of perfusion fraction and blood velocity dispersion estimates in intravoxel incoherent motion (IVIM) imaging, using joint analysis of flow‐compensated and non‐flow‐compensated motion‐encoded MRI data. A double diffusion encoding sequence capable of switching between flow‐compensated and non‐flow‐compensated encoding modes was implemented. In vivo brain data were collected in eight healthy volunteers and processed using the joint analysis. Simulations were used to compare the performance of the proposed analysis method with conventional IVIM analysis. With flow compensation, strong rephasing was observed for the in vivo data, approximately cancelling the IVIM effect. The joint analysis yielded physiologically reasonable perfusion fraction maps. Estimated perfusion fractions were 2.43 ± 0.81% in gray matter, 1.81 ± 0.90% in deep gray matter, and 1.64 ± 0.72% in white matter (mean ± SD, n = 8). Simulations showed improved accuracy and precision when using joint analysis of flow‐compensated and non‐flow‐compensated data, compared with conventional IVIM analysis. Double diffusion encoding with flow compensation was feasible for in vivo imaging of the perfusion fraction in the brain. The strong rephasing implied that blood flowing through the cerebral microvascular system was closer to the ballistic limit than the diffusive limit. © 2016 The Authors NMR in Biomedicine published by John Wiley & Sons Ltd.     

Intravoxel incoherent motion imaging kinetics during chemoradiotherapy for human papillomavirus‐associated squamous cell carcinoma of the oropharynx: preliminary results from a prospective pilot study

This study aims to identify the temporal kinetics of intravoxel incoherent motion (IVIM) MRI in patients with human papillomavirus‐associated (HPV+) oropharyngeal squamous cell carcinoma. Patients were enrolled under an Institutional Review Board (IRB)‐approved protocol as part of an ongoing prospective clinical trial. All patients underwent two MRI studies: a baseline scan before chemoradiotherapy and a mid‐treatment scan 3–4 weeks after treatment initiation. Parametric maps representing pure diffusion coefficient (D), pseudo‐diffusion coefficient (D*), perfusion fraction (f) and apparent diffusion coefficient (ADC) were generated. The Mann–Whitney U‐test was used to assess the temporal variation of IVIM metrics. Bayesian quadratic discriminant analysis (QDA) was used to evaluate the extent to which mid‐treatment changes in IVIM metrics could be combined to predict sites that would achieve complete response (CR) in multivariate analysis. Thirty‐one patients were included in the final analysis with 59 lesions. Pretreatment ADC and D values of the CR lesions (n = 19) were significantly lower than those of non‐CR lesions (n = 33). Mid‐treatment ADC, D and f values were significantly higher (p < 0.0001) than pretreatment values for all lesions. Each increase in normalized ΔADC of size 0.1 yielded a 1.45‐fold increase in the odds of CR (p < 0.0003), each increase in normalized ΔD of size 0.1 yielded a 1.53‐fold increase in the odds of CR (p < 0.0002), and each unit increase in Δf yielded a 2.29‐fold increase in the odds of CR (p < 0.02). Combined ΔD and ΔADC were integrated into a multivariate prediction model and attained an AUC of 0.87 (95% confidence interval: 0.79, 0.96), as well as a sensitivity of 0.63, specificity of 0.85 and accuracy of 0.78, under leave‐one‐out cross‐validation. In conclusion, IVIM is feasible and potentially useful in the prediction and assessment of the early response of HPV+ oropharyngeal squamous cell carcinoma to chemoradiotherapy. Copyright © 2015 John Wiley & Sons, Ltd.     

The effect of noise and lipid signals on determination of Gaussian and non‐Gaussian diffusion parameters in skeletal muscle

This work characterizes the effect of lipid and noise signals on muscle diffusion parameter estimation in several conventional and non‐Gaussian models, the ultimate objectives being to characterize popular fat suppression approaches for human muscle diffusion studies, to provide simulations to inform experimental work and to report normative non‐Gaussian parameter values. The models investigated in this work were the Gaussian monoexponential and intravoxel incoherent motion (IVIM) models, and the non‐Gaussian kurtosis and stretched exponential models. These were evaluated via simulations, and in vitro and in vivo experiments. Simulations were performed using literature input values, modeling fat contamination as an additive baseline to data, whereas phantom studies used a phantom containing aliphatic and olefinic fats and muscle‐like gel. Human imaging was performed in the hamstring muscles of 10 volunteers. Diffusion‐weighted imaging was applied with spectral attenuated inversion recovery (SPAIR), slice‐select gradient reversal and water‐specific excitation fat suppression, alone and in combination. Measurement bias (accuracy) and dispersion (precision) were evaluated, together with intra‐ and inter‐scan repeatability. Simulations indicated that noise in magnitude images resulted in <6% bias in diffusion coefficients and non‐Gaussian parameters (α, K), whereas baseline fitting minimized fat bias for all models, except IVIM. In vivo, popular SPAIR fat suppression proved inadequate for accurate parameter estimation, producing non‐physiological parameter estimates without baseline fitting and large biases when it was used. Combining all three fat suppression techniques and fitting data with a baseline offset gave the best results of all the methods studied for both Gaussian diffusion and, overall, for non‐Gaussian diffusion. It produced consistent parameter estimates for all models, except IVIM, and highlighted non‐Gaussian behavior perpendicular to muscle fibers (α ~ 0.95, K ~ 3.1). These results show that effective fat suppression is crucial for accurate measurement of non‐Gaussian diffusion parameters, and will be an essential component of quantitative studies of human muscle quality.     

An intravoxel oriented flow model for diffusion‐weighted imaging of the kidney

By combining intravoxel incoherent motion (IVIM) and diffusion tensor imaging (DTI) we introduce a new diffusion model called intravoxel oriented flow (IVOF) that accounts for anisotropy of diffusion and the flow‐related signal. An IVOF model using a simplified apparent flow fraction tensor (IVOF_f) is applied to diffusion‐weighted imaging of human kidneys. The kidneys of 13 healthy volunteers were examined on a 3 T scanner. Diffusion‐weighted images were acquired with six b values between 0 and 800 s/mm² and 30 diffusion directions. Diffusivity and flow fraction were calculated for different diffusion models. The Akaike information criterion was used to compare the model fit of the proposed IVOF_f model to IVIM and DTI. In the majority of voxels the proposed IVOF_f model with a simplified apparent flow fraction tensor performs better than IVIM and DTI. Mean diffusivity is significantly higher in DTI compared with models that account for the flow‐related signal. The fractional anisotropy of diffusion is significantly reduced when flow fraction is considered to be anisotropic. Anisotropy of the apparent flow fraction tensor is significantly higher in the renal medulla than in the cortex region. The IVOF_f model describes diffusion‐weighted data in the human kidney more accurately than IVIM or DTI. The apparent flow fraction in the kidney proved to be anisotropic.     

Differentiating the histologic grades of gliomas preoperatively using amide proton transfer‐weighted (APTW) and intravoxel incoherent motion MRI

The purpose of this work was to investigate the diagnostic performance of amide proton transfer‐weighted (APTW) and intravoxel incoherent motion (IVIM) magnetic resonance imaging (MRI) in the preoperative grading of gliomas. Fifty‐one patients with suspected gliomas were recruited and underwent a preoperative MRI examination that included APTW and IVIM sequences. All cases were confirmed by postsurgical histopathology. APTW signal intensity, true diffusion coefficient (D), perfusion fraction (f) and pseudo‐diffusion coefficient (D*) were applied to assess the solid tumor component and contralateral normal‐appearing white matter. The relative APTW signal intensity (rAPTW) was also used. Independent‐sample and paired‐sample t‐tests were used to compare differences in MRI parameters between low‐grade glioma (LGG) and high‐grade glioma (HGG) groups. The diagnostic performance was assessed with the receiver operating characteristic curve. Twenty‐six patients were pathologically diagnosed with LGG and 25 were diagnosed with HGG. APTW, rAPTW and f values were significantly higher (all p < 0.001), whereas D values were significantly lower (p < 0.001) in the HGG group than in the LGG group. There was no significant difference between D* values for the two groups. rAPTW had an area under the curve (AUC) of 0.957, with a sensitivity of 100% and a specificity of 84.6%, followed by APTW, f, D and D*. The combined use of APTW and IVIM showed the best diagnostic performance, with an AUC of 0.986. In conclusion, APTW and IVIM, as two promising supplementary sequences for routine MRI, could be valuable in differentiating LGGs from HGGs.     

Assessment of diffusion parameters by intravoxel incoherent motion MRI in head and neck squamous cell carcinoma

The objectives of this study were to assess the diffusion parameters derived from intravoxel incoherent motion (IVIM) MRI in head and neck squamous cell carcinoma (HNSCC) and to investigate the agreement between different methods of tumor delineation and two numerical methods to extract the perfusion fraction f. Thirty‐seven untreated patients with histopathologically confirmed primary HNSCC were included retrospectively in the study. The entire volume of the primary tumor was outlined on diffusion‐weighted images using co‐registered morphological images as a guide to the tumor location. Apparent diffusion coefficient (ADC) and IVIM diffusion parameters were estimated considering the largest tumor section as well as the entire tumor volume. A bi‐exponential fit was implemented to extract f, D (pure diffusion coefficient) and D* (pseudo‐diffusion coefficient). A second simplified method, based on an asymptotic extrapolation, was used to determine f. The agreement between ADC and IVIM diffusion parameters derived from the delineation of single and multiple slices, and between the two f estimations, was assessed by Bland–Altman plots. The inter‐slice variability of ADC and IVIM diffusion parameters was evaluated. The Kruskal–Wallis test was used to investigate whether the tumor location had a statistically significant influence on the values of the parameters. Comparing the tumor delineation methods, a better accordance was found for ADC and D, with a mean percentage difference of less than 2%. Larger discrepancies were found for f and D*, with mean differences of 4.5% and 5.5%, respectively. When comparing the two f estimation methods, small mean differences were found (<3.5%), suggesting that the two methods may be considered as equivalent for the assessment of f in our patient population. The observed ADC and IVIM diffusion parameters were dependent on the anatomic site of the lesion, carcinoma of the nasopharynx showing more homogeneous and dissimilar estimations than other HNSCCs. Copyright © 2013 John Wiley & Sons, Ltd.     

Interstitial fluid pressure correlates with intravoxel incoherent motion imaging metrics in a mouse mammary carcinoma model

The effective delivery of a therapeutic drug to the core of a tumor is often impeded by physiological barriers, such as the interstitial fluid pressure (IFP). There are a number of therapies that can decrease IFP and induce tumor vascular normalization. However, a lack of a noninvasive means to measure IFP hinders the utilization of such a window of opportunity for the maximization of the treatment response. Thus, the purpose of this study was to investigate the feasibility of using intravoxel incoherent motion (IVIM) diffusion parameters as noninvasive imaging biomarkers for IFP. Mice bearing the 4T1 mammary carcinoma model were studied using diffusion‐weighted imaging (DWI), immediately followed by wick‐in‐needle IFP measurement. Voxelwise analysis was conducted with a conventional monoexponential diffusion model, as well as a biexponential model taking IVIM into account. There was no significant correlation of IFP with either the median apparent diffusion coefficient from the monoexponential model (r = 0.11, p = 0.78) or the median tissue diffusivity from the biexponential model (r = 0.30, p = 0.44). However, IFP was correlated with the median pseudo‐diffusivity (D_p) of apparent vascular voxels (r = 0.76, p = 0.02) and with the median product of the perfusion fraction and pseudo‐diffusivity (f_pD_p) of apparent vascular voxels (r = 0.77, p = 0.02). Although the effect of IVIM in tumors has been reported previously, to our knowledge, this study represents the first direct comparison of IVIM metrics with IFP, with the results supporting the feasibility of the use of IVIM DWI metrics as noninvasive biomarkers for tumor IFP. Copyright © 2011 John Wiley & Sons, Ltd.     

Correlative assessment of tumor microcirculation using contrast‐enhanced perfusion MRI and intravoxel incoherent motion diffusion‐weighted MRI: is there a link between them?

The purpose of this study was to correlate intravoxel incoherent motion (IVIM) imaging with classical perfusion‐weighted MRI metrics in human gliomas. Parametric images for slow diffusion coefficient (D), fast diffusion coefficient (D*), and fractional perfusion‐related volume (f) in patients with high‐grade gliomas were generated. Maps of F_p (plasma flow), v_p (vascular plasma volume), PS (permeability surface–area product), v_e (extravascular, extracellular volume), E (extraction ratio), k_e (influx ratio into the interstitium), and t_c (vascular transit time) from dynamic contrast‐enhanced (DCE) and dynamic susceptibility contrast‐enhanced (DSC) MRI were also generated. A region‐of‐interest analysis on the contralateral healthy white matter and on the tumor areas was performed and the extracted parameter values were tested for any significant differences among tumor grades or any correlations. Only f could be significantly correlated to DSC‐derived v_p and t_c in healthy brain tissue. Concerning the tumor regions, F_p was significantly positively correlated with D* and inversely correlated with f in DSC measurements. The D*, f, and f × D* values in the WHO grade III gliomas were non‐significantly different from those in the grade IV gliomas. There was a trend to significant negative correlations between f and PS as well as between f × D* and k_e in DCE experiments. Presumably due to different theoretical background, tracer properties and modeling of the tumor vasculature in the IVIM theory, there is no clearly evident link between D*, f and DSC‐ and DCE‐derived metrics. Copyright © 2014 John Wiley & Sons, Ltd.     

Measurement of distinctive features of cortical spreading depolarizations with different MRI contrasts

Growing clinical evidence suggests critical involvement of spreading depolarizations (SDs) in the pathophysiology of neurological disorders such as migraine and stroke. MRI provides powerful tools to detect and assess co‐occurring cerebral hemodynamic and cellular changes during SDs. This study reports the feasibility and advantages of two MRI scans, based on balanced steady‐state free precession (b‐SSFP) and diffusion‐weighted multi‐spin‐echo (DT2), heretofore unexplored for monitoring SDs. These were compared with gradient‐echo MRI. SDs were induced by KCl application in rat brain. Known for high SNR, the T₂‐ and T₁‐based b‐SSFP contrast was hypothesized to provide higher spatiotemporal specificity than ‐based gradient‐echo scanning. DT2 scanning was designed to provide simultaneous T₂ and apparent diffusion coefficient (ADC) measurements, thus enabling combined quantitative assessment of hemodynamic and cellular changes during SDs. Procedures were developed to automate identification of SD‐induced responses in all the scans. These responses were analyzed to determine detection sensitivity and temporal characteristics of signals from each scanning method. Cluster analysis was performed to elucidate unique temporal patterns for each contrast. All scans allowed detection of SD‐induced responses. b‐SSFP scans showed significantly larger relative intensity changes, narrower peak widths and greater spatial specificity compared with gradient‐echo MRI. SD‐induced effects on ADC, calculated from DT2 scans, showed the most pronounced signal changes, displaying about 20% decrease, as against 10–15% signal increases observed with b‐SSFP and gradient‐echo scanning. Cluster analysis revealed additional temporal sub‐patterns, such as an initial dip on gradient‐echo scans and temporally shifted T₂ and proton density changes in DT2 data. To summarize, b‐SSFP and DT2 scanning provide distinct information on SDs compared with gradient‐echo MRI. DT2 scanning, with its potential to simultaneously provide cellular and hemodynamic information, can offer unique information on the inter‐relationship between these processes in pathologic brain, which may improve monitoring of spreading depolarizations in (pre)clinical settings. Copyright © 2015 John Wiley & Sons, Ltd.     

Multiparametric classification of sub‐acute ischemic stroke recovery with ultrafast diffusion, 23Na,and MPIO‐labeled stem cell MRI at 21.1 T

MRI leverages multiple modes of contrast to characterize stroke. High‐magnetic‐field systems enhance the performance of these MRI measurements. Previously, we have demonstrated that individually sodium and stem cell tracking metrics are enhanced at ultrahigh field in a rat model of stroke, and we have developed robust single‐scan diffusion‐weighted imaging approaches that utilize spatiotemporal encoding (SPEN) of the apparent diffusion coefficient (ADC) for these challenging field strengths. Here, we performed a multiparametric study of middle cerebral artery occlusion (MCAO) biomarker evolution focusing on comparison of these MRI biomarkers for stroke assessment during sub‐acute recovery in rat MCAO models at 21.1 T. T₂‐weighted MRI was used as the benchmark for identification of the ischemic lesion over the course of the study. The number of MPIO‐induced voids measured by gradient‐recalled echo, the SPEN measurement of ADC, and ²³Na MRI values were determined in the ischemic area and contralateral hemisphere, and relative performances for stroke classification were compared by receiver operator characteristic analysis. These measurements were associated with unique time‐dependent trajectories during stroke recovery that changed the sensitivity and specificity for stroke monitoring during its evolution. Advantages and limitations of these contrasts, and the use of ultrahigh field for multiparametric stroke assessment, are discussed.     

Increased brain perfusion contrast with T2‐prepared intravoxel incoherent motion (T2prep IVIM) MRI

The feasibility to measure brain perfusion using intravoxel incoherent motion (IVIM) MRI has been reported recently with currently clinically available technology. The method is intrinsically local and quantitative, but is contaminated by partial volume effects with cerebrospinal fluid (CSF). Signal from CSF can be suppressed by a 180° inversion recovery (180°‐IR) magnetization preparation, but this also leads to strong suppression of blood and brain tissue signal. Here, we take advantage of the different T₂ relaxations of blood and brain relative to CSF, and implement a T₂‐prepared IVIM (T2prep IVIM) inversion recovery acquisition, which permits a recovery of between 43% and 57% of arterial and venous blood magnetization at excitation time compared with the theoretical recovery of between 27% and 30% with a standard 180°‐IR. We acquired standard IVIM (IVIM), T2prep IVIM and dynamic susceptibility contrast (DSC) images at 3 T using a 32‐multichannel receiver head coil in eight patients with known large high‐grade brain tumors. We compared the contrast and contrast‐to‐noise ratio obtained in the corresponding cerebral blood volume images quantitatively, as well as subjectively by two neuroradiologists. Our findings suggest that quantitative cerebral blood volume contrast and contrast‐to‐noise ratio, as well as subjective lesion detection, contrast quality and diagnostic confidence, are increased with T2prep IVIM relative to IVIM and DSC. Copyright © 2014 John Wiley & Sons, Ltd.     

Dynamic intravoxel incoherent motion imaging of skeletal muscle at rest and after exercise

The purpose of this work was to demonstrate the feasibility of intravoxel incoherent motion imaging (IVIM) for non‐invasive quantification of perfusion and diffusion effects in skeletal muscle at rest and following exercise. After IRB approval, eight healthy volunteers underwent diffusion‐weighted MRI of the forearm at 3 T and eight different b values between 0 and 500 s/mm² with a temporal resolution of 57 s per dataset. Dynamic images were acquired before and after a standardized handgrip exercise. Diffusion (D) and pseudodiffusion (D*) coefficients as well as the perfusion fraction (F_P) were measured in regions of interest in the flexor digitorum superficialis and profundus (FDS/FDP), brachioradialis, and extensor carpi radialis longus and brevis muscles by using a multi‐step bi‐exponential analysis in MATLAB. Parametrical maps were calculated voxel‐wise. Differences in D, D*, and F_P between muscle groups and between time points were calculated using a repeated measures analysis of variance with post hoc Bonferroni tests. Mean values and standard deviations at rest were the following: D*, 28.5 ± 11.4 × 10^?3 mm²/s; F_P, 0.03 ± 0.01; D, 1.45 ± 0.09 × 10^?3 mm²/s. Changes of IVIM parameters were clearly visible on the parametrical maps. In the FDS/FDP, D* increased by 289 ± 236% (p < 0.029), F_P by 138 ± 58% (p < 0.01), and D by 17 ± 9% (p < 0.01). A significant increase of IVIM parameters could also be detected in the brachioradialis muscle, which however was significantly lower than in the FDS/FDP. After 20 min, all parameters were still significantly elevated in the FDS/FDP but not in the brachioradialis muscle compared with the resting state. The IVIM approach allows simultaneous quantification of muscle perfusion and diffusion effects at rest and following exercise. It may thus provide a useful alternative to other non‐invasive methods such as arterial spin labeling. Possible fields of interest for this technique include perfusion‐related muscle diseases, such as peripheral arterial occlusive disease. Copyright © 2014 John Wiley & Sons, Ltd.     

Classification of parotid gland tumors by using multimodal MRI and deep learning

Various MRI sequences have shown their potential to discriminate parotid gland tumors, including but not limited to T₂‐weighted, postcontrast T₁‐weighted, and diffusion‐weighted images. In this study, we present a fully automatic system for the diagnosis of parotid gland tumors by using deep learning methods trained on multimodal MRI images. We used a two‐dimensional convolution neural network, U‐Net, to segment and classify parotid gland tumors. The U‐Net model was trained with transfer learning, and a specific design of the batch distribution optimized the model accuracy. We also selected five combinations of MRI contrasts as the input data of the neural network and compared the classification accuracy of parotid gland tumors. The results indicated that the deep learning model with diffusion‐related parameters performed better than those with structural MR images. The performance results (n = 85) of the diffusion‐based model were as follows: accuracy of 0.81, 0.76, and 0.71, sensitivity of 0.83, 0.63, and 0.33, and specificity of 0.80, 0.84, and 0.87 for Warthin tumors, pleomorphic adenomas, and malignant tumors, respectively. Combining diffusion‐weighted and contrast‐enhanced T₁‐weighted images did not improve the prediction accuracy. In summary, the proposed deep learning model could classify Warthin tumor and pleomorphic adenoma tumor but not malignant tumor.     

Multi‐centre reproducibility of diffusion MRI parameters for clinical sequences in the brain

The purpose of this work was to assess the reproducibility of diffusion imaging, and in particular the apparent diffusion coefficient (ADC), intra‐voxel incoherent motion (IVIM) parameters and diffusion tensor imaging (DTI) parameters, across multiple centres using clinically available protocols with limited harmonization between sequences. An ice–water phantom and nine healthy volunteers were scanned across fives centres on eight scanners (four Siemens 1.5T, four Philips 3T). The mean ADC, IVIM parameters (diffusion coefficient D and perfusion fraction f) and DTI parameters (mean diffusivity MD and fractional anisotropy FA), were measured in grey matter, white matter and specific brain sub‐regions. A mixed effect model was used to measure the intra‐ and inter‐scanner coefficient of variation (CV) for each of the five parameters. ADC, D, MD and FA had a good intra‐ and inter‐scanner reproducibility in both grey and white matter, with a CV ranging between 1% and 7.4%; mean 2.6%. Other brain regions also showed high levels of reproducibility except for small structures such as the choroid plexus. The IVIM parameter f had a higher intra‐scanner CV of 8.4% and inter‐scanner CV of 24.8%. No major difference in the inter‐scanner CV for ADC, D, MD and FA was observed when analysing the 1.5T and 3T scanners separately. ADC, D, MD and FA all showed good intra‐scanner reproducibility, with the inter‐scanner reproducibility being comparable or faring slightly worse, suggesting that using data from multiple scanners does not have an adverse effect compared with using data from the same scanner. The IVIM parameter f had a poorer inter‐scanner CV when scanners of different field strengths were combined, and the parameter was also affected by the scan acquisition resolution. This study shows that the majority of diffusion MRI derived parameters are robust across 1.5T and 3T scanners and suitable for use in multi‐centre clinical studies and trials. © 2015 The Authors NMR in Biomedicine Published by John Wiley & Sons Ltd.     

Automated pixel‐wise brain tissue segmentation of diffusion‐weighted images via machine learning

The diffusion‐weighted (DW) MR signal sampled over a wide range of b‐values potentially allows for tissue differentiation in terms of cellularity, microstructure, perfusion, and T₂ relaxivity. This study aimed to implement a machine learning algorithm for automatic brain tissue segmentation from DW‐MRI datasets, and to determine the optimal sub‐set of features for accurate segmentation. DWI was performed at 3 T in eight healthy volunteers using 15 b‐values and 20 diffusion‐encoding directions. The pixel‐wise signal attenuation, as well as the trace and fractional anisotropy (FA) of the diffusion tensor, were used as features to train a support vector machine classifier for gray matter, white matter, and cerebrospinal fluid classes. The datasets of two volunteers were used for validation. For each subject, tissue classification was also performed on 3D T₁‐weighted data sets with a probabilistic framework. Confusion matrices were generated for quantitative assessment of image classification accuracy in comparison with the reference method. DWI‐based tissue segmentation resulted in an accuracy of 82.1% on the validation dataset and of 82.2% on the training dataset, excluding relevant model over‐fitting. A mean Dice coefficient (DSC) of 0.79 ± 0.08 was found. About 50% of the classification performance was attributable to five features (i.e. signal measured at b‐values of 5/10/500/1200 s/mm² and the FA). This reduced set of features led to almost identical performances for the validation (82.2%) and the training (81.4%) datasets (DSC = 0.79 ± 0.08). Machine learning techniques applied to DWI data allow for accurate brain tissue segmentation based on both morphological and functional information.     

Correlation of tumor characteristics derived from DCE‐MRI and DW‐MRI with histology in murine models of breast cancer

The purpose of this work was to determine the relationship between the apparent diffusion coefficient (ADC, from diffusion‐weighted (DW) MRI), the extravascular, extracellular volume fraction (v_e, from dynamic contrast‐enhanced (DCE) MRI), and histological measurement of the extracellular space fraction. Athymic nude mice were injected with either human epidermal growth factor receptor 2 positive (HER2+) BT474 (n = 15) or triple negative MDA‐MB‐231 (n = 20) breast cancer cells, treated with either Herceptin (n = 8), Abraxane (low dose n = 7, high dose n = 6), or saline (n = 7 for each cell line), and imaged using DW‐ and DCE‐MRI before, during, and after treatment. After the final imaging acquisition, the tissue was resected and evaluated by histological analysis. H&E‐stained central slices were scanned using a digital brightfield microscope and evaluated with thresholding techniques to calculate the extracellular space. For both BT474 and MDA‐MB‐231, the median ADC of the central slice exhibited a significantly positive correlation with the corresponding central slice extracellular space as measured by H&E (p = 0.03, p < 0.01, respectively). Median v_e calculated from the central slice showed differing results between the two cell lines. For BT474, a significant correlation between v_e and extracellular space was calculated (p = 0.02), while MDA‐MB‐231 tumors did not demonstrate a significant correlation (p = 0.64). Additionally, there was no correlation discovered between ADC and v_e with either whole tumor analysis or central slice analysis (p > 0.05). While ADC correlates well with the histologically determined fraction of extracellular space, these data add to the growing body of literature that suggests that v_e derived from DCE‐MRI is not a reliable biomarker of extracellular space for a range of physiological conditions. Copyright © 2015 John Wiley & Sons, Ltd.