Harmonizing 1.5T/3T Diffusion Weighted MRI through Development of Deep Learning Stabilized Microarchitecture Estimators.

Whole-Body Diffusion-Weighted Magnetic Resonance Imaging Is a New Sensitive Non-Invasive Method for Staging and Monitoring Chemotherapy Response of Malignant Lymphoma Including Indolent Lymphoma Compared with FDG-PET/CT

Evaluation of early regression index as response predictor in cervical cancer: A retrospective study on T2 and DWI MR images

Multi-tissue constrained spherical deconvolution for improved analysis of multi-shell diffusion MRI data

Diffusion weighted magnetic resonance imaging (DW-MRI) captures tissue microarchitecture at millimeter scale. With recent advantages in data sharing, large-scale multi-site DW-MRI datasets are being made available for multi-site studies. However, DW-MRI suffers from measurement variability (e.g., inter- and intra-site variability, hardware performance, and sequence design), which consequently yields inferior performance on multi-site and/or longitudinal diffusion studies. In this study, we propose a novel, deep learning-based method to harmonize DW-MRI signals for a more reproducible and robust estimation of microstructure. Our method introduces a data-driven scanner-invariant regularization scheme to model a more robust fiber orientation distribution function (FODF) estimation. We study the Human Connectome Project (HCP) young adults test-retest group as well as the MASiVar dataset (with inter- and intra-site scan/rescan data). The 8th order spherical harmonics coefficients are employed as data representation. The results show that the proposed harmonization approach maintains higher angular correlation coefficients (ACC) with the ground truth signals (0.954 versus 0.942), while achieves higher consistency of FODF signals for intra-scanner data (0.891 versus 0.826), as compared with the baseline supervised deep learning scheme. Furthermore, the proposed data-driven framework is flexible and potentially applicable to a wider range of data harmonization problems in neuroimaging.

Constrained spherical deconvolution (CSD) of diffusion‐weighted MRI (DW‐MRI) is a popular analysis method that extracts the full white matter (WM) fiber orientation density function (fODF) in the living human brain, noninvasively. It assumes that the DW‐MRI signal on the sphere can be represented as the spherical convolution of a single‐fiber response function (RF) and the fODF, and recovers the fODF through the inverse operation. CSD approaches typically require that the DW‐MRI data is sampled shell‐wise, and estimate the RF in a purely spherical manner using spherical basis functions, such as spherical harmonics (SH), disregarding any radial dependencies. This precludes analysis of data acquired with nonspherical sampling schemes, for example, Cartesian sampling. Additionally, nonspherical sampling can also arise due to technical issues, for example, gradient nonlinearities, resulting in a spatially dependent bias of the apparent tissue densities and connectivity information. Here, we adopt a compact model for the RFs that also describes their radial dependency. We demonstrate that the proposed model can accurately predict the tissue response for a wide range of b‐values. On shell‐wise data, our approach provides fODFs and tissue densities indistinguishable from those estimated using SH. On Cartesian data, fODF estimates and apparent tissue densities are on par with those obtained from shell‐wise data, significantly broadening the range of data sets that can be analyzed using CSD. In addition, gradient nonlinearities can be accounted for using the proposed model, resulting in much more accurate apparent tissue densities and connectivity metrics.

MR diagnostic of postoperative recurrent cholesteatomas is difficult. Our purpose was to compare multishot fast spin-echo periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) diffusion-weighted MR imaging (DWI) with array spatial sensitivity encoding technique (ASSET) single-shot echo-planar DWI and late postgadolinium T1-weighted MR imaging for the detection of postoperative recurrent middle ear cholesteatomas with a 3T imaging unit. Thirty-five patients with suggested postoperative recurrent middle ear cholesteatoma underwent 3T MR imaging with PROPELLER DWI, ASSET echo-planar DWI, and late postgadolinium T1-weighted MR imaging. Three radiologists (2 seniors, 1 fellow) analyzed unlabeled images for visualization of recurrence. Interobserver and intraobserver agreement was assessed by using the Cohen kappa statistic test. Sensitivity, specificity, and predictive value were assessed for the 3 observers. Nineteen recurrent cholesteatomas were diagnosed. PROPELLER interobserver agreement was very good (1, 0.89, 0.89) among the 3 observers. Intraobserver agreement between PROPELLER and T1-weighted imaging was very good to moderate (0.88, 0.57, 0.58). PROPELLER DWI provided less interobserver variability than other sequences, and the best sensitivity, specificity, and predictive value. On a 3T imaging unit, multishot fast spin-echo PROPELLER DWI allows an easier detection of postoperative recurrent middle ear cholesteatoma than T1-weighted imaging by reducing artifacts and by its better contrast. DWI with PROPELLER is diagnostically robust and accurate.

This prospective clinical study assesses the feasibility of training a deep neural network (DNN) for intravoxel incoherent motion (IVIM) model fitting to diffusion-weighted MRI (DW-MRI) data and evaluates its performance. In May 2011, 10 male volunteers (age range, 29-53 years; mean, 37) underwent DW-MRI of the upper abdomen on 1.5T and 3.0T MR scanners. Regions of interest in the left and right liver lobe, pancreas, spleen, renal cortex, and renal medulla were delineated independently by 2 readers. DNNs were trained for IVIM model fitting using these data; results were compared to least-squares and Bayesian approaches to IVIM fitting. Intraclass correlation coefficients (ICCs) were used to assess consistency of measurements between readers. Intersubject variability was evaluated using coefficients of variation (CVs). The fitting error was calculated based on simulated data, and the average fitting time of each method was recorded. DNNs were trained successfully for IVIM parameter estimation. This approach was associated with high consistency between the 2 readers (ICCs between 50% and 97%), low intersubject variability of estimated parameter values (CVs between 9.2 and 28.4), and the lowest error when compared with least-squares and Bayesian approaches. Fitting by DNNs was several orders of magnitude quicker than the other methods, but the networks may need to be retrained for different acquisition protocols or imaged anatomical regions. DNNs are recommended for accurate and robust IVIM model fitting to DW-MRI data. Suitable software is available for download.

Diffusion-weighted (DW) magnetic resonance imaging (MRI) is a non-invasive imaging method, which can be used to investigate neural tracts in the white matter (WM) of the brain. Significant partial volume effects (PVEs) are present in the DW signal due to relatively large voxel sizes. These PVEs can be caused by both non-WM tissue, such as gray matter (GM) and cerebrospinal fluid (CSF), and by multiple non-parallel WM fiber populations. High angular resolution diffusion imaging (HARDI) methods have been developed to correctly characterize complex WM fiber configurations, but to date, many of the HARDI methods do not account for non-WM PVEs. In this work, we investigated the isotropic PVEs caused by non-WM tissue in WM voxels on fiber orientations extracted with constrained spherical deconvolution (CSD). Experiments were performed on simulated and real DW-MRI data. In particular, simulations were performed to demonstrate the effects of varying the diffusion weightings, signal-to-noise ratios (SNRs), fiber configurations, and tissue fractions. Our results show that the presence of non-WM tissue signal causes a decrease in the precision of the detected fiber orientations and an increase in the detection of false peaks in CSD. We estimated 35–50% of WM voxels to be affected by non-WM PVEs. For HARDI sequences, which typically have a relatively high degree of diffusion weighting, these adverse effects are most pronounced in voxels with GM PVEs. The non-WM PVEs become severe with 50% GM volume for maximum spherical harmonics orders of 8 and below, and already with 25% GM volume for higher orders. In addition, a low diffusion weighting or SNR increases the effects. The non-WM PVEs may cause problems in connectomics, where reliable fiber tracking at the WM–GM interface is especially important. We suggest acquiring data with high diffusion-weighting 2500–3000 s/mm2, reasonable SNR (~30) and using lower SH orders in GM contaminated regions to minimize the non-WM PVEs in CSD.

High angular resolution diffusion imaging (HARDI) methods have enabled the reconstruction of complex spin diffusion profiles in central nervous system white matter through diffusion-weighted MRI. For recovery of the underlying fiber orientations, conventional spherical deconvolution techniques based on spherical harmonics typically have difficulty producing fiber orientation distributions (FODs) that simultaneously satisfy the physical constraints of being real, symmetric, and non-negative. In this work, we propose a novel approach for HARDI reconstruction that is guaranteed to generate FODs satisfying these constraints. By using a meshed representation of the unit sphere, we formulate the spherical deconvolution as a convex optimization problem and compute the solution using a projected gradient descent algorithm. Flexible regularization is also included in our method to allow for tuning the sharpness of the reconstructed FOD. In our experiments, we present simulated results to examine the effects of varying the regularization parameters, and we illustrate the robustness of our method by applying it to several biological data sets to reconstruct known white matter fiber geometry.

As a multi-fiber tracking model, the constrained spherical deconvolution (CSD) model is widely used in the field of fiber reconstruction. The CSD model has shown good reconstruction capabilities for crossing fibers in low anisotropy regions, which can achieve more accurate results in terms of brain fiber reconstruction. However, the current fiber tracking algorithms based on the CSD model have a few drawbacks in the selection of tracking strategies, especially in the certain crossing regions, which may lead to isotropic diffusion signals, premature termination of fibers, high computational complexity, and low efficiency. In this study, we proposed the fiber tracking method with adaptive selection of peak direction based on CSD model, called FTASP_CSD, for fiber reconstruction. The method first filters the fiber orientation distribution (FOD) peak threshold and eliminates peak directions lower than the set threshold. Secondly, a priority strategy is used to implement direction selection, and the tracking direction is adaptively adjusted according to the overall shape and needs of the FOD. Through dynamic selection of the maximum peak direction, the second maximum peak direction and the nearest peak direction, the tracking direction that best matches the true fiber direction is found. This method not only ensures spatial consistency, but also avoids the influence of stray peaks in the FOD that may be introduced by imaging noise on the fiber tracking direction. Experimental results on simulation and in vivo data show that the fiber bundles tracked by the FTASP_CSD method have a much smoother in the overall visual effect than the state-of-the-art methods. The fiber bundles tracked in the region of crossing or bifurcating fibers are more complete. This improves the angular resolution of the recognition of fiber crossings and lays a foundation for further in-depth research on fiber tracking technology.

Distribution of collateral fibers in the monkey cervical spinal cord detected with diffusion-weighted magnetic resonance imaging

Purpose. To evaluate diffusion weighted magnetic resonance imaging (DWI) in liver and liver cancers during and following conformal radiotherapy (RT). To determine the feasibility of using changes in apparent diffusion coefficients (ADC) as a potential surrogate for tumour control or normal tissue injury. Materials and methods. Patients on a six-fraction conformal liver RT protocol underwent DW-MRI at the time of treatment planning, during RT (week one and two) and one month following RT. Diffusion weighted MR images were acquired in exhale breath hold, using b-values of 0 and 600. Regions of interest (ROIs) corresponding to maximal tumour dose, high-dose peri-tumour liver, irradiated normal liver, non-irradiated liver, and spleen were analyzed on ADC maps. Results. Eleven patients (four hepatocellular carcinoma, five liver metastases, two cholangiocarcinoma) were evaluated. The baseline median tumour ADC of 1.56×10−3mm2/sec increased to 1.89×10-3mm2/sec at RT week one, to1.91x10-3mm2/sec during week two and to 2.01×10-3mm2/sec at one month following treatment (p < 0.0001). Early increases in mean ADC were correlated with higher dose and sustained tumour response, whereas RECIST and volume changes on T2 images were not. Peri-tumour mean ADC also increased, from 1.40×10−3mm2/sec (baseline) to 1.55×10−3mm2/sec (RT week 2) and 1.64 × 10−3mm2/sec (follow-up). Small ADC changes were seen in the irradiated liver, and no significant changes were seen in the un-irradiated liver. Conclusions. Changes in tumour ADC were seen during RT. Larger increases were correlated with higher doses and increased likelihood of response.

The goal of diffusion-weighted magnetic resonance imaging (DWI) is to infer the structural connectivity of an individual subject’s brain in vivo. To statistically study the variability and differences between normal and abnormal brain connectomes, a mathematical model of the neural connections is required. In this paper, we represent the brain connectome as a Riemannian manifold, which allows us to model neural connections as geodesics. This leads to the challenging problem of estimating a Riemannian metric that is compatible with the DWI data, i.e., a metric such that the geodesic curves represent individual fiber tracts of the connectomics. We reduce this problem to that of solving a highly nonlinear set of partial differential equations (PDEs) and study the applicability of convolutional encoder-decoder neural networks (CEDNNs) for solving this geometrically motivated PDE. Our method achieves excellent performance in the alignment of geodesics with white matter pathways and tackles a long-standing issue in previous geodesic tractography methods: the inability to recover crossing fibers with high fidelity. Code is available at https://github.com/aarentai/Metric-Cnn-3D-IPMI .

To evaluate the efficacy of targeted prostate biopsy using magnetic resonance imaging (MRI) and to characterize clinicopathologic features of tumors detected with targeted prostate biopsy in men with previous negative prostate biopsy results. We prospectively studied 87 patients with a persistently increasing level of serum prostate-specific antigen (PSA), at least one previous set of negative 12-core prostate biopsies, and normal digital rectal examination. All patients were examined with combined T2-weighted and diffusion-weighted MRI before undergoing the prostate biopsy. Prostate biopsy was performed using transrectal ultrasonography-guided standard 12 cores plus targeted biopsy to suspicious region(s) as identified on T2 images on their MRI. Of a total of 87 cases, 82 (94.2%) patients had suspicious lesion(s) on their MRI. Of these 82 patients, 46 (56.0%) patients had prostate cancer (PCA) as determined by the biopsy. The patients with PCA showed a significantly higher incidence of having suspicious lesion(s) (the anterior or apex) on MRI than the patients without PCA (P<0.05). On analysis by dividing all biopsy cores into the targeted cores and standard cores, PCA was found in 149/518 (28.8%) MRI-targeted cores and in 32/903 (3.6%) standard cores (P=0.012) Of 43 patients who underwent radical prostatectomy, 37 (86.0%) patients were detected with PCA located at the anterior or apex portion of the prostate. For tumor characteristics according to tumor locations, there was no significant correlation between tumor location and Gleason scores or pathologic stage. Our data suggest that a MRI-targeted prostate biopsy after prostate MRI might be considered for the identification of cancer foci and the detection of PCA, for patients with a previous negative standard prostate biopsy result despite a persistently elevated PSA value.

Background and Objective: We explored dynamic changes in the choroid plexus (CP) in patients with relapsing-remitting multiple sclerosis (RRMS) and assessed its relationship with chronic lesion expansion and atrophy in various brain compartments. Methods: Fifty-seven RRMS patients were annually assessed for a minimum of 48 months with 3D FLAIR, pre- and post-contrast 3D T1 and diffusion-weighted magnetic resonance imaging (MRI). The CP was manually segmented at baseline and last follow-up. Results: The volume of CP significantly increased by 1.4% annually. However, the extent of CP enlargement varied considerably among individuals (ranging from −3.6 to 150.8 mm3 or −0.2% to 6.3%). The magnitude of CP enlargement significantly correlated with central (r = 0.70, p < 0.001) and total brain atrophy (r = −0.57, p < 0.001), white (r = −0.61, p < 0.001) and deep grey matter atrophy (r = −0.60, p < 0.001). Progressive CP enlargement was significantly associated with the volume and extent of chronic lesion expansion (r = 0.60, p < 0.001), but not with the number or volume of new lesions. Conclusion: This study provides evidence of progressive CP enlargement in patients with RRMS. Our findings also demonstrate that enlargement of the CP volume is linked to the expansion of chronic lesions and neurodegeneration of periventricular white and grey matter in RRMS patients.