Characterization of breast lesions using multi-parametric diffusion MRI and machine learning

Objective. To investigate quantitative imaging markers based on parameters from two diffusion-weighted imaging (DWI) models, continuous-time random-walk (CTRW) and intravoxel incoherent motion (IVIM) models, for characterizing malignant and benign breast lesions by using a machine learning algorithm.Approach. With IRB approval, 40 women with histologically confirmed breast lesions (16 benign, 24 malignant) underwent DWI with 11b-values (50 to 3000 s/mm2) at 3T. Three CTRW parameters,Dm,α, andβand three IVIM parametersDdiff,Dperf, andfwere estimated from the lesions. A histogram was generated and histogram features of skewness, variance, mean, median, interquartile range; and the value of the 10%, 25% and 75% quantiles were extracted for each parameter from the regions-of-interest. Iterative feature selection was performed using the Boruta algorithm that uses the Benjamin Hochberg False Discover Rate to first determine significant features and then to apply the Bonferroni correction to further control for false positives across multiple comparisons during the iterative procedure. Predictive performance of the significant features was evaluated using Support Vector Machine, Random Forest, Naïve Bayes, Gradient Boosted Classifier (GB), Decision Trees, AdaBoost and Gaussian Process machine learning classifiers.Main Results. The 75% quantile, and median ofDm; 75% quantile off;mean, median, and skewness ofβ;kurtosis ofDperf; and 75% quantile ofDdiffwere the most significant features. The GB differentiated malignant and benign lesions with an accuracy of 0.833, an area-under-the-curve of 0.942, and an F1 score of 0.87 providing the best statistical performance (p-value < 0.05) compared to the other classifiers.Significance. Our study has demonstrated that GB with a set of histogram features from the CTRW and IVIM model parameters can effectively differentiate malignant and benign breast lesions.

Simultaneous multi-segment (SMSeg) EPI over multiple focal regions

Objective.This study aimed at developing a simultaneous multi-segment (SMSeg) imaging technique using a two-dimensional (2D) RF pulse in conjunction with echo planar imaging (EPI) to image multiple focal regions.Approach.The SMSeg technique leveraged periodic replicates of the excitation profile of a 2D RF pulse to simultaneously excite multiple focal regions at different locations. These locations were controlled by rotating and scaling transmit k-space trajectories. The resulting multiple isolated focal regions were projected into a composite 'slice' for display. GRAPPA-based parallel imaging was incorporated into SMSeg by taking advantage of coil sensitivity variations in both the phase-encoded and slice-selection directions. The SMSeg technique was implemented at 3 T in a single-shot gradient-echo EPI sequence and demonstrated in a phantom and human brains for both anatomic imaging and functional imaging.Main results.In both the phantom and the human brain, SMSeg images from three focal regions were simultaneously acquired. SMSeg imaging enabled up to a six-fold acceleration in parallel imaging without causing appreciable residual aliasing artifacts when compared with a conventional gradient-echo EPI sequence with the same acceleration factor. In the functional imaging experiment, BOLD activations associated with a visuomotor task were simultaneously detected in two non-coplanar segments (each with a size of 240 × 30 mm²), corresponding to visual and motor cortices, respectively.Significance.Our study has demonstrated that SMSeg imaging can be a viable method for studying multiple focal regions simultaneously.

In-plane simultaneous multisegment imaging using a 2D RF pulse

PURPOSE

To develop an in-plane simultaneous multisegment (IP-SMS) imaging technique using a 2D-RF pulse and to demonstrate its ability to achieve high spatial resolution in EPI while reducing image distortion.

METHODS

The proposed IP-SMS technique takes advantage of periodic replicates of the excitation profile of a 2D-RF pulse to simultaneously excite multiple segments within a slice. These segments were acquired over a reduced FOV and separated using a joint GRAPPA reconstruction by leveraging virtual coils that combined the physical coil sensitivity and 2D-RF pulse spatial response. Two excitations were used with complementary spatial response profiles to adequately cover a full FOV, producing a full-FOV image that had the benefits of reduced FOV with high spatial resolution and reduced distortion. The IP-SMS technique was implemented in a diffusion-weighted single-shot EPI sequence. Experimental demonstrations were performed on a phantom and healthy human brain.

RESULTS

In the phantom experiment, IP-SMS enabled a four-fold acceleration using an eight-channel coil without causing residual aliasing artifacts. In the human brain experiment, diffusion-weighted images with high in-plane resolution (1 × 1 mm² ) and substantially reduced image distortion were obtained in all imaging planes in comparison with a commercial diffusion-weighted EPI sequence. The capability of IP-SMS for contiguous whole-brain coverage was also demonstrated.

CONCLUSION

The proposed IP-SMS technique can realize the benefits of reduced-FOV imaging while achieving a full-FOV coverage with good image quality and time efficiency.

Evaluation of a fractional-order calculus diffusion model and bi-parametric VI-RADS for staging and grading bladder urothelial carcinoma

OBJECTIVES

To evaluate the feasibility of high b-value diffusion-weighted imaging (DWI) for distinguishing non-muscle-invasive bladder cancer (NMIBC) from muscle-invasive bladder cancer (MIBC) and low- from high-grade bladder urothelial carcinoma using a fractional-order calculus (FROC) model as well as a combination of FROC DWI and bi-parametric Vesical Imaging-Reporting and Data System (VI-RADS).

METHODS

Fifty-eight participants with bladder urothelial carcinoma were included in this IRB-approved prospective study. Diffusion-weighted images, acquired with 16 b-values (0-3600 s/mm²), were analyzed using the FROC model. Three FROC parameters, D, β, and μ, were used for delineating NMIBC from MIBC and for tumor grading. A receiver operating characteristic (ROC) analysis was performed based on the individual FROC parameters and their combinations, followed by comparisons with apparent diffusion coefficient (ADC) and bi-parametric VI-RADS based on T2-weighted images and DWI.

RESULTS

D and μ were significantly lower in the MIBC group than in the NMIBC group (p = 0.001 for each), and D, β, and μ all exhibited significantly lower values in the high- than in the low-grade tumors (p ≤ 0.011). The combination of D, β, and μ produced the highest specificity (85%), accuracy (78%), and the area under the ROC curve (AUC, 0.782) for distinguishing NMIBC and MIBC, and the best sensitivity (89%), specificity (86%), accuracy (88%), and AUC (0.892) for tumor grading, all of which outperformed the ADC. The combination of FROC parameters with bi-parametric VI-RADS improved the AUC from 0.859 to 0.931.

CONCLUSIONS

High b-value DWI with a FROC model is useful in distinguishing NMIBC from MIBC and grading bladder tumors.

KEY POINTS

• Diffusion parameters derived from a FROC diffusion model may differentiate NMIBC from MIBC and low- from high-grade bladder urothelial carcinomas. • Under the condition of a moderate sample size, higher AUCs were achieved by the FROC parameters D (0.842) and μ (0.857) than ADC (0.804) for bladder tumor grading with p ≤ 0.046. • The combination of the three diffusion parameters from the FROC model can improve the specificity over ADC (85% versus 67%, p = 0.031) for distinguishing NMIBC and MIBC and enhance the performance of bi-parametric VI-RADS.

Condensation Artifact

During the ongoing COVID-19 pandemic, an artifact with hyperintense signal was observed on the brain images of a number of patients or research subjects, particularly those with heavy body weight and/or increased respiratory rate. The artifact was primarily seen on 3D or 2D sagittal or coronal T2-weighted images, although it occasionally also appeared in the axial plane. It manifested as a bright spot or a cluster of bright spots at similar locations, superior or lateral superior to the skull. This artifact was found to be caused by condensed water droplet(s) in the head coil as a consequence of the altered moisture flow pattern associated with each exhalation due to the mask on the patient. We call this artifact condensation artifact. Several strategies have been proposed to prevent or resolve the artifact. LEVEL OF EVIDENCE: 5 TECHNICAL EFFICACY: Stage 1.

Quartile histogram assessment of glioma malignancy using high b-value diffusion MRI with a continuous-time random-walk model

The purpose of this study is to investigate the feasibility of using a continuous-time random-walk (CTRW) diffusion model, together with a quartile histogram analysis, for assessing glioma malignancy by probing tissue heterogeneity as well as cellularity. In this prospective study, 91 patients (40 females, 51 males) with histopathologically proven gliomas underwent MRI at 3 T. The cohort included 42 grade II (GrII), 19 grade III (GrIII) and 29 grade IV (GrIV) gliomas. Echo-planar diffusion-weighted imaging was conducted using 17 b-values (0-4000 s/mm² ). Three CTRW model parameters, including an anomalous diffusion coefficient D_m , and two parameters related to temporal and spatial diffusion heterogeneity α and β, respectively, were obtained. The mean parameter values within the tumor regions of interest (ROIs) were computed by utilizing the first quartile of the histograms as well as the full ROI for comparison. A Bonferroni-Holm-corrected Mann-Whitney U-test was used for the group comparisons. Individual and combinations of the CTRW parameters were evaluated for the characterization of gliomas with a receiver operating characteristic analysis. All first-quartile mean CTRW parameters yielded significant differences (p-values < 0.05) between pair-wise comparisons of GrII (D_m : 1.14 ± 0.37 μm² /ms; α: 0.904 ± 0.03, β: 0.913 ± 0.06), GrIII (D_m : 0.88 ± 0.21 μm² /ms; α: 0.888 ± 0.01, β: 0.857 ± 0.06) and GrIV gliomas (D_m : 0.73 ± 0.22 μm² /ms; α: 0.878 ± 0.01; β: 0.791 ± 0.07). The highest sensitivity, specificity, accuracy and area-under-the-curve of using the combinations of the first-quartile parameters were 84.2%, 78.5%, 75.4% and 0.76 for GrII and GrIII classification; 86.2%, 89.4%, 75% and 0.76 for GrIII and GrIV classification; and 86.2%, 85.7%, 84.5% and 0.90 for GrII and GrIV classification, respectively. Quartile-based analysis produced higher accuracy and area-under-the-curve than the full ROI-based analysis in all classifications. The CTRW diffusion model, together with a quartile-based histogram analysis, offers a new way for probing tumor structural heterogeneity at a subvoxel level, and has potential for in vivo assessment of glioma malignancy to complement histopathology.

In vivo assessment of Lauren classification for gastric adenocarcinoma using diffusion MRI with a fractional order calculus model

OBJECTIVES

To evaluate the performance of a fractional order calculus (FROC) diffusion model for imaging-based assessment of Lauren classification in gastric adenocarcinoma.

METHODS

In this study, 43 patients (15 females, 28 males) with gastric adenocarcinoma underwent MRI at 1.5 T. According to pathology-based Lauren classification, 10 patients had diffuse-type, 20 had intestinal-type, and 13 had mixed-type lesions. The diffuse and mixed types were combined as diffuse-and-mixed type to be differentiated from the intestinal type using diffusion MRI. Diffusion-weighted images were acquired by using eleven b-values (0-2000 s/mm²). Three FROC model parameters comprising diffusion coefficient D, intravoxel diffusion heterogeneity β, and a microstructural quantity μ, together with a conventional apparent diffusion coefficient (ADC), were estimated. The mean parameter values in the tumour were computed by using a percentile histogram analysis. Individual or linear combinations of the mean parameters in the tumour were used to differentiate the diffuse-and-mixed type from the intestinal type using descriptive statistics and receiver operating characteristic (ROC) analyses.

RESULTS

Significant differences were observed between diffuse-and-mixed-type and intestinal-type lesions in D (0.99 ± 0.20 μm²/ms vs. 1.11 ± 0.23 μm²/ms; p = 0.036), β (0.37 ± 0.08 vs. 0.43 ± 0.11; p = 0.043), μ (7.92 ± 2.79 μm vs. 9.87 ± 1.52 μm; p = 0.038), and ADC (0.81 ± 0.34 μm²/ms vs. 0.96 ± 0.19 μm²/ms; p = 0.033). Among the individual parameters, μ produced the largest area under the ROC curve (0.739). The combinations of (D, β, μ) and (β and μ) produced the best overall performance with a sensitivity of 0.739, specificity of 0.750, accuracy of 0.744, and area under the curve of 0.793 (95% confidence interval: 0.657-0.929).

CONCLUSION

Diffusion MRI with the FROC model holds promise for non-invasive assessment of Lauren classification for gastric adenocarcinoma.

KEY POINTS

• High b-value diffusion MRI with a FROC model that is sensitive to tissue microstructures can differentiate the diffuse-and-mixed type from intestinal type of gastric adenocarcinoma. • The combination of FROC parameters produced the best result for distinguishing the diffuse-and-mixed type from the intestinal type with an area under the receiver operating characteristic curve of 0.793. • The FROC model parameters, individually or conjointly, hold promise for repeated, non-invasive evaluations of gastric adenocarcinoma at various time points throughout disease progression or regression to complement conventional Lauren classification.

Magnetic resonance imaging with submillisecond temporal resolution

PURPOSE

To demonstrate an MRI technique-Submillisecond Periodic Event Encoded Dynamic Imaging (SPEEDI)-for capturing cyclic dynamic events with submillisecond temporal resolution.

METHODS

The SPEEDI technique is based on an FID or an echo signal in which each time point in the signal is used to sample a distinct k-space raster, followed by repeated FIDs or echoes to produce the remaining k-space data in each k-space raster. All acquisitions are synchronized with a cyclic event, resulting in a set of time-resolved images of the cyclic event with a temporal resolution determined by the dwell time. In SPEEDI, spatial encoding is accomplished by phase encoding. The SPEEDI technique was demonstrated in two experiments at 3 T to (1) visualize fast-changing electric currents that mimicked the waveform of an action potential, and (2) characterize rapidly decaying eddy currents in an MRI system, with a temporal resolution of 0.2 ms and 0.4 ms, respectively. In both experiments, compressed sensing was incorporated to reduce the scan times. Phase difference maps related to the dynamics of electric currents or eddy currents were then obtained.

RESULTS

In the first experiment, time-resolved phase maps resulting from the action potential-mimicking current waveform were successfully obtained and agreed well with theoretical calculations (normalized RMS error = 0.07). In the second experiment, spatially resolved eddy current phase maps revealed time constants (27.1 ± 0.2 ms, 41.1 ± 3.5 ms, and 34.8 ± 0.7 ms) that matched well with those obtained from an established method using point sources (26.4 ms, 41.2 ms and 34.8 ms). For both experiments, phase maps from fully sampled and compressed-sensing-accelerated k-space data exhibited a high structural similarity (> 0.8) despite a two-fold to three-fold acceleration.

CONCLUSIONS

We have illustrated that SPEEDI can provide submillisecond temporal resolution. This capability will likely lead to future exploration of ultrafast, cyclic biomedical processes using MRI.

Fractional Calculus Models of Magnetic Resonance Phenomena: Relaxation and Diffusion

Applications of fractional calculus in magnetic resonance imaging (MRI) have increased over the last twenty years. From the mathematical, computational, and biophysical perspectives, fractional calculus provides new tools for describing the complexity of biological tissues (cells, organelles, membranes and macromolecules). Specifically, fractional order models capture molecular dynamics (transport, rotation, and vibration) by incorporating power law convolution kernels into the time and space derivatives appearing in the equations that govern nuclear magnetic resonance (NMR) phenomena. Hence, it is natural to expect fractional calculus models of relaxation and diffusion to be applied to problems in NMR and MRI. Early studies considered the fractal dimensions of multi-scale materials in the non-linear growth of the mean squared displacement, assumed power-law decays of the spectral density, and suggested stretched exponential signal relaxation to describe non-Gaussian behavior. Subsequently, fractional order generalization of the Bloch, and Bloch-Torrey equations were developed to characterize NMR (and MRI) relaxation and diffusion. However, even for simple geometries, analytical solutions of fractional order equations in time and space are difficult to obtain, and predictions of the corresponding changes in image contrast are not always possible. Currently, a multifaceted approach using coarse graining, simulation, and accelerated computation is being developed to identify 'imaging' biomarkers of disease. This review surveys the principal fractional order models used to describe NMR and MRI phenomena, identifies connections and limitations, and finally points to future applications of the approach.

High-Spatial-Resolution Diffusion MRI in Parkinson Disease: Lateral Asymmetry of the Substantia Nigra

Background Motor symptoms in Parkinson disease (PD) have exhibited lateral asymmetry, suggesting asymmetric neuronal loss in the substantia nigra (SN). Diffusion MRI may be able to help confirm tissue microstructural alterations in the substantia nigra to probe for the presence of asymmetry. Purpose To investigate lateral asymmetry in the SN of patients with PD by using diffusion MRI with both Gaussian and non-Gaussian models. Materials and Methods In this cross-sectional study conducted from March 2015 to March 2017, 27 participants with PD and 27 age-matched healthy control (HC) participants, all right handed, underwent MRI at 3.0 T. High-spatial-resolution diffusion images were acquired with a reduced field of view by using seven b values up to 3000 sec/mm². A continuous-time random-walk (CTRW) non-Gaussian diffusion model was used to produce anomalous diffusion coefficient (D_m) and temporal (α) and spatial (β) diffusion heterogeneity indexes followed by a Gaussian diffusion model to yield an apparent diffusion coefficient (ADC). Individual or linear combinations of diffusion parameters in the SN were unilaterally and bilaterally compared between the PD and HC groups. Results In the bilateral comparison between the PD and HC groups, differences were observed in β (0.67 ± 0.06 [standard deviation] vs 0.64 ± 0.04, respectively; P = .016), ADC (0.48 μm²/msec ± 0.08 vs 0.53 μm²/msec ± 0.06, respectively; P = .03), and the combination of CTRW parameters (P = .02). In the unilateral comparison, differences were observed in all diffusion parameters on the left SN (P < .03), but not on the right (P > .20). In a receiver operating characteristic (ROC) analysis to delineate left SN abnormality in PD, the combination of D_m, α, and β produced the best sensitivity (sensitivity, 0.78); the combination of D_m and β produced the best specificity (specificity, 0.85); and the combination of α and β produced the largest area under the ROC curve (area under the ROC curve, 0.73). Conclusion These results suggest that quantitative diffusion MRI is sensitive to brain tissue changes in participants with Parkinson disease and provide evidence of substantia nigra lateral asymmetry in this disease. © RSNA, 2019 Online supplemental material is available for this article.

A fractional motion diffusion model for a twice-refocused spin-echo pulse sequence

The purpose of this study was to develop an analytical expression for a fractional motion (FM) diffusion model to characterize diffusion-induced signal attenuation in a twice-refocused spin-echo (TRSE) sequence that is resilient to eddy currents, and to demonstrate its applicability to human brain imaging in vivo. Based on the FM theory, which provides a unified statistical description for Langevin motions, the diffusion-weighted (DW) MR signal was measured with a TRSE sequence that balances the concomitant gradients. The analytical expression was fitted to a set of DW images acquired with 14 b-values (0-4000 s/mm² ) from a total of 10 healthy human subjects at 3 T, yielding three FM parameter maps based on anomalous diffusion coefficient D_{φ, ψ} , diffusion increment variance φ, and diffusion correlation ψ, respectively. These parameters were used to characterize different brain regions in gray matter (GM), white matter (WM), and cerebrospinal fluid. The analytical expression for the TRSE-based FM model accurately described diffusion signal attenuation in healthy brain tissues at high b-values. TRSE's robustness against eddy currents was illustrated by comparing results from an expression for a conventional Stejskal-Tanner sequence. The TRSE-based FM model also produced consistent GM-WM contrast (p < 0.01) across all brain regions studied, whereas the consistency was not observed with the Stejskal-Tanner-based FM model. This new analytical expression is expected to enable further investigations to probe tissue structures by exploiting anomalous diffusion properties without being hindered by eddy-current perturbations at high b-values.

Capturing complexity of the diffusion-weighted MR signal decay

Diffusion-weighted MRI (dMRI) is a key component of clinical radiology. When analyzing diffusion-weighted images, radiologists often seek to infer microscopic tissue structure through measurements of the diffusion coefficient, D₀ (mm²/s). This multi-scale problem is framed by the creation of diffusion models of signal decay based on physical laws, histological structure, and biophysical constraints. The purpose of this paper is to simplify the model building process by focusing on the observed decay in the effective diffusion coefficient as a function of diffusion weighting (b-value), D(b), that is often observed in complex biological tissues. We call this approach the varying diffusion curvature (VDC) model. Since this is a heuristic model, the exact functional form of this decay is not important, so here we examine a simple exponential function, D(b) = D₀exp(-bD₁), where D₀ and D₁ capture aspects of hindered and restricted diffusion, respectively. As an example of the potential of the VDC model, we applied it to dMRI data collected from normal and diseased human brain tissue using Stejskal-Tanner diffusion gradient pulses. In order to illustrate the connection between D₀ and D₁ and the sub-voxel structure we also analyzed dMRI data from families of Sephadex beads selected with increasing tortuosity. Finally, we applied the VDC model to dMRI simulations of nested muscle fiber phantoms whose permeability, atrophy, and fiber size distribution could be changed. These results demonstrate that the VDC model is sensitive to sub-voxel tissue structure and composition (porosity, tortuosity, and permeability), hence can capture tissue complexity in a manner that could be easily applied in clinical dMRI.

White matter structural differences in OSA patients experiencing residual daytime sleepiness with high CPAP use: a non-Gaussian diffusion MRI study

OBJECTIVES

To investigate factors associated with residual sleepiness in patients who were highly adherent to continuous positive airway pressure (CPAP). Nocturnal inactivity, comorbidities, concomitant medications, and, in particular, white matter (WM) differences using diffusion magnetic resonance imaging (MRI) were explored using a continuous-time random-walk (CTRW) model.

METHODS

Twenty-seven male patients (30-55 years of age) with obstructive sleep apnea (OSA) received CPAP as the only treatment (CPAP ≥ 6 h/night) for at least 30 days. Based on the Psychomotor Vigilance Task (PVT) results, participants were divided into a non-sleepy group (lapses ≤ 5; n = 18) and a sleepy group (lapses > 5; n = 9). Mean nocturnal inactivity (sleep proxy) was measured using actigraphy for one week. Diffusion-weighted imaging (DWI) with high b-values, as well as diffusion tensor imaging (DTI), was performed on a 3 T MRI scanner. The DWI dataset was analyzed using the CTRW model that yielded three parameters: temporal diffusion heterogeneity α, spatial diffusion heterogeneity β, and an anomalous diffusion coefficient D_m. The differences in α, β, and D_m between the two groups were investigated by a whole-brain analysis using tract-based spatial statistics (TBSS), followed by a regional analysis on individual fiber tracts using a standard parcellation template. Results from the CTRW model were compared with those obtained from DTI. The three CTRW parameters were also correlated with the clinical assessment scores, Epworth Sleepiness Scale (ESS), PVT lapses, and PVT mean reaction time (MRT) in specific fiber tracts.

RESULTS

There were no differences between groups in mean sleep duration, comorbidities, and the number or type of medications, including alerting and sedating medications. In the whole-brain DWI analysis, the sleepy group showed higher α (17.27% of the WM voxels) and D_m (17.14%) when compared to the non-sleepy group (P < 0.05), whereas no significant difference in β was observed. In the regional fiber analysis, the sleepy and non-sleepy groups showed significant differences in α, β, or their combinations in a total of 12 fiber tracts; whereas similar differences were not observed in DTI parameters, when age was used as a covariate. Additionally, moderate to strong correlations between the CTRW parameters (α, β, or D_m) and the sleepiness assessment scores (ESS, PVT lapses, or PVT MRT) were observed in specific fiber tracts (|R| = 0.448-0.654, P = 0.0003-0.019).

CONCLUSIONS

The observed differences in the CTRW parameters between the two groups indicate that WM alterations can be a possible mechanism to explain reversible versus residual sleepiness observed in OSA patients with identical high level of CPAP use. The moderate to strong correlations between the CTRW parameters and the clinical scores suggest the possibility of developing objective and quantitative imaging markers to complement clinical assessment of OSA patients.

A fractional motion diffusion model for grading pediatric brain tumors

Objectives

To demonstrate the feasibility of a novel fractional motion (FM) diffusion model for distinguishing low- versus high-grade pediatric brain tumors; and to investigate its possible advantage over apparent diffusion coefficient (ADC) and/or a previously reported continuous-time random-walk (CTRW) diffusion model.

Materials and methods

With approval from the institutional review board and written informed consents from the legal guardians of all participating patients, this study involved 70 children with histopathologically-proven brain tumors (30 low-grade and 40 high-grade). Multi-b-value diffusion images were acquired and analyzed using the FM, CTRW, and mono-exponential diffusion models. The FM parameters, D_fm, φ, ψ (non-Gaussian diffusion statistical measures), and the CTRW parameters, D_m, α, β (non-Gaussian temporal and spatial diffusion heterogeneity measures) were compared between the low- and high-grade tumor groups by using a Mann-Whitney-Wilcoxon U test. The performance of the FM model for differentiating between low- and high-grade tumors was evaluated and compared with that of the CTRW and the mono-exponential models using a receiver operating characteristic (ROC) analysis.

Results

The FM parameters were significantly lower (p < 0.0001) in the high-grade (D_fm: 0.81 ± 0.26, φ: 1.40 ± 0.10, ψ: 0.42 ± 0.11) than in the low-grade (D_fm: 1.52 ± 0.52, φ: 1.64 ± 0.13, ψ: 0.67 ± 0.13) tumor groups. The ROC analysis showed that the FM parameters offered better specificity (88% versus 73%), sensitivity (90% versus 82%), accuracy (88% versus 78%), and area under the curve (AUC, 93% versus 80%) in discriminating tumor malignancy compared to the conventional ADC. The performance of the FM model was similar to that of the CTRW model.

Conclusions

Similar to the CTRW model, the FM model can improve differentiation between low- and high-grade pediatric brain tumors over ADC.

•

The fractional motion (FM) diffusion model was applied to pediatric brain tumors.

•

The FM model parameters can be sensitive to tissue microstructures.

•

The FM model outperforms the mono-exponential diffusion model.

•

The FM model performs similarly to the continuous-time random-walk (CTRW) model.

•

Our results challenge those from recent biophysics studies in cell cultures.

Differentiating low- and high-grade pediatric brain tumors using a continuous-time random-walk diffusion model at high b-values

PURPOSE

To demonstrate that a continuous-time random-walk (CTRW) diffusion model can improve diagnostic accuracy of differentiating low- and high-grade pediatric brain tumors.

METHODS

Fifty-four children with histopathologically confirmed brain tumors underwent diffusion MRI scans at 3Twith 12 b-values (0-4000 s/mm(2) ). The diffusion imageswere fit to a simplified CTRW model to extract anomalous diffusion coefficient, Dm , and temporal and spatial heterogeneity parameters, α and β, respectively. Using histopathology results as reference, a k-means clustering algorithm and a receiver operating characteristic (ROC) analysis were employed to determine the sensitivity, specificity, and diagnostic accuracy of the CTRW parameters in differentiating tumor grades.

RESULTS

Significant differences between the low- and high-grade tumors were observed in the CTRW parameters (p-values<0.001). The k-means analysis showed that the combination of three CTRW parameters produced higher diagnostic accuracy (85% vs. 75%) and specificity (83% vs. 54%) than the apparent diffusion coefficient (ADC) from a mono-exponential model. The ROC analysis revealed that any combination of the CTRW parameters gave a larger area under the curve (0.90-0.96) than using ADC (0.80).

CONCLUSION

With its sensitivity to intravoxel heterogeneity, the simplified CTRW model is useful for non-invasive grading of pediatric brain tumors, particularly when surgical biopsy is not feasible. Magn Reson Med 76:1149-1157, 2016. © 2015 Wiley Periodicals, Inc.

Brain white matter changes in CPAP-treated obstructive sleep apnea patients with residual sleepiness

PURPOSE

To investigate white matter (WM) structural alterations using diffusion tensor imaging (DTI) in obstructive sleep apnea (OSA) patients, with or without residual sleepiness, following adherent continuous positive airway pressure (CPAP) treatment. Possible quantitative relationships were explored between the DTI metrics and two clinical assessments of somnolence.

MATERIALS AND METHODS

Twenty-nine male patients (30-55 years old) with a confirmed diagnosis of OSA were recruited. The patients were treated with CPAP therapy only. The Psychomotor Vigilance Task (PVT) and Epworth Sleepiness Scale (ESS) were performed after CPAP treatment and additionally administered at the time of the magnetic resonance imaging (MRI) scan. Based on the PVT results, the patients were divided into a nonsleepy group (lapses ≤5) and a sleepy group (lapses >5). DTI was performed at 3T, followed by an analysis using tract-based spatial statistics (TBSS) to investigate the differences in fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (λ₁ ), and radial diffusivity (λ₂₃ ) between the two groups.

RESULTS

A higher MD (P < 0.05) was observed in the sleepy group than the nonsleepy group in the whole-brain TBSS analysis in the WM. The increased MD (17.8% of the fiber tracts; P < 0.05) was caused primarily by an elevated λ₂₃ . Axial diffusivity (λ₁ ) exhibited no significant difference (P > 0.17). The alterations in FA or MD of individual fiber tracts occurred mainly in the internal/external capsule, corona radiata, corpus callosum, and sagittal stratum regions. The FA and MD values correlated with the PVT and ESS assessments from all patients (R ≥ 0.517, P < 0.05).

CONCLUSION

Global and regional WM alterations, as revealed by DTI, can be a possible mechanism to explain why OSA patients with high levels of CPAP use can have differing responses to treatment. Compromised myelin sheath, indicated by increased radial diffusivity, can be involved in the underlying WM changes. Evidence level: 1 J. MAGN. RESON. IMAGING 2017;45:1371-1378.

Differentiation of Low- and High-Grade Gliomas Using High b-Value Diffusion Imaging with a Non-Gaussian Diffusion Model

BACKGROUND AND PURPOSE

Imaging-based tumor grading is highly desirable but faces challenges in sensitivity, specificity, and diagnostic accuracy. A recently proposed diffusion imaging method by using a fractional order calculus model offers a set of new parameters to probe not only the diffusion process itself but also intravoxel tissue structures, providing new opportunities for noninvasive tumor grading. This study aimed to demonstrate the feasibility of using the fractional order calculus model to differentiate low- from high-grade gliomas in adult patients and illustrate its improved performance over a conventional diffusion imaging method using ADC (or D).

MATERIALS AND METHODS

Fifty-four adult patients (18-70 years of age) with histology-proved gliomas were enrolled and divided into low-grade (n = 24) and high-grade (n = 30) groups. Multi-b-value diffusion MR imaging was performed with 17 b-values (0-4000 s/mm(2)) and was analyzed by using a fractional order calculus model. Mean values and SDs of 3 fractional order calculus parameters (D, β, and μ) were calculated from the normal contralateral thalamus (as a control) and the tumors, respectively. On the basis of these values, the low- and high-grade glioma groups were compared by using a Mann-Whitney U test. Receiver operating characteristic analysis was performed to assess the performance of individual parameters and the combination of multiple parameters for low- versus high-grade differentiation.

RESULTS

Each of the 3 fractional order calculus parameters exhibited a statistically higher value (P ≤ .011) in the low-grade than in the high-grade gliomas, whereas there was no difference in the normal contralateral thalamus (P ≥ .706). The receiver operating characteristic analysis showed that β (area under the curve = 0.853) produced a higher area under the curve than D (0.781) or μ (0.703) and offered a sensitivity of 87.5%, specificity of 76.7%, and diagnostic accuracy of 82.1%.

CONCLUSIONS

The study demonstrated the feasibility of using a non-Gaussian fractional order calculus diffusion model to differentiate low- and high-grade gliomas. While all 3 fractional order calculus parameters showed statistically significant differences between the 2 groups, β exhibited a better performance than the other 2 parameters, including ADC (or D).

A Diffusion Tensor Imaging Study on White Matter Abnormalities in Patients with Type 2 Diabetes Using Tract-Based Spatial Statistics

BACKGROUND AND PURPOSE

Patients with type 2 diabetes mellitus have considerably higher risk of developing cognitive impairment and dementia. WM changes in these patients have been reported. Our aim was to demonstrate that gradual and continuous WM change and the associated cognitive decline in patients with type 2 diabetes mellitus can be captured by DTI parameters, which can be used to complement neuropsychological test scores in identifying patients with type 2 diabetes mellitus with and without mild cognitive impairment.

MATERIALS AND METHODS

Forty-two patients with type 2 diabetes mellitus, divided into a group with mild cognitive impairment (n = 20) and a group with normal cognition (n = 22), were enrolled with age-, sex-, and education-matched healthy controls (n = 26). 3T DTI followed by Tract-Based Spatial Statistics analysis was used to investigate the differences in fractional anisotropy, mean diffusivity, axial diffusivity (λ1), and radial diffusivity (λ23) among the groups. A receiver operating characteristic analysis assessed the performance of DTI parameters for separating the 2 groups with type 2 diabetes mellitus.

RESULTS

The whole-brain Tract-Based Spatial Statistics analysis revealed that 7.3% and 24.9% of the WM exhibited decreased fractional anisotropy and increased mean diffusivity (P < .05), respectively, between the diabetes mellitus with mild cognitive impairment and the diabetes mellitus with normal cognition groups, while considerably larger WM regions showed fractional anisotropy (36.6%) and mean diffusivity (58.8%) changes between the diabetes mellitus with mild cognitive impairment and the healthy control groups. These changes were caused primarily by an elevated radial diffusivity observed in the patients with diabetes mellitus with mild cognitive impairment. Radial diffusivity also exhibited subtle but statistically significant changes between the diabetes mellitus with normal cognition and the healthy control groups. Analyses on individual fiber tracts showed pronounced fractional anisotropy reduction and mean diffusivity elevation in regions related to cognitive functions. The receiver operating characteristic analysis on the right cingulum (hippocampus) showed that fractional anisotropy produced a larger area under the curve (0.832) than mean diffusivity (0.753) for separating mild cognitive impairment from normal cognition among patients with type 2 diabetes mellitus. When fractional anisotropy was combined with mean diffusivity, the area under the curve was further improved to 0.857.

CONCLUSIONS

DTI parameters can show a substantial difference between patients with type 2 diabetes mellitus with and without mild cognitive impairment, suggesting their potential use as an imaging marker for detecting cognitive decline in patients with type 2 diabetes mellitus. More important, DTI parameters may capture gradual and continuous WM changes that can be associated with early stages of cognitive decline in patients with type 2 diabetes mellitus before they can be diagnosed clinically by using conventional neuropsychological tests.

Differentiating low- and high-grade pediatric brain tumors using a continuous-time random-walk diffusion model at high b-values

PURPOSE

To demonstrate that a continuous-time random-walk (CTRW) diffusion model can improve diagnostic accuracy of differentiating low- and high-grade pediatric brain tumors.

METHODS

Fifty-four children with histopathologically confirmed brain tumors underwent diffusion MRI scans at 3Twith 12 b-values (0-4000 s/mm(2) ). The diffusion imageswere fit to a simplified CTRW model to extract anomalous diffusion coefficient, Dm , and temporal and spatial heterogeneity parameters, α and β, respectively. Using histopathology results as reference, a k-means clustering algorithm and a receiver operating characteristic (ROC) analysis were employed to determine the sensitivity, specificity, and diagnostic accuracy of the CTRW parameters in differentiating tumor grades.

RESULTS

Significant differences between the low- and high-grade tumors were observed in the CTRW parameters (p-values<0.001). The k-means analysis showed that the combination of three CTRW parameters produced higher diagnostic accuracy (85% vs. 75%) and specificity (83% vs. 54%) than the apparent diffusion coefficient (ADC) from a mono-exponential model. The ROC analysis revealed that any combination of the CTRW parameters gave a larger area under the curve (0.90-0.96) than using ADC (0.80).

CONCLUSION

With its sensitivity to intravoxel heterogeneity, the simplified CTRW model is useful for non-invasive grading of pediatric brain tumors, particularly when surgical biopsy is not feasible. Magn Reson Med 76:1149-1157, 2016. © 2015 Wiley Periodicals, Inc.

A statistical fMRI model for differential T2* contrast incorporating T1 and T2* of gray matter

Relaxation parameter estimation and brain activation detection are two main areas of study in magnetic resonance imaging (MRI) and functional magnetic resonance imaging (fMRI). Relaxation parameters can be used to distinguish voxels containing different types of tissue whereas activation determines voxels that are associated with neuronal activity. In fMRI, the standard practice has been to discard the first scans to avoid magnetic saturation effects. However, these first images have important information on the MR relaxivities for the type of tissue contained in voxels, which could provide pathological tissue discrimination. It is also well-known that the voxels located in gray matter (GM) contain neurons that are to be active while the subject is performing a task. As such, GM MR relaxivities can be incorporated into a statistical model in order to better detect brain activation. Moreover, although the MR magnetization physically depends on tissue and imaging parameters in a nonlinear fashion, a linear model is what is conventionally used in fMRI activation studies. In this study, we develop a statistical fMRI model for Differential T2(*) ConTrast Incorporating T1 and T2(*) of GM, so-called DeTeCT-ING Model, that considers the physical magnetization equation to model MR magnetization; uses complex-valued time courses to estimate T1 and T2(*) for each voxel; then incorporates gray matter MR relaxivities into the statistical model in order to better detect brain activation, all from a single pulse sequence by utilizing the first scans.

A statistical examination of SENSE image reconstruction via an isomorphism representation

In magnetic resonance imaging, the parallel acquisition of subsampled spatial frequencies from an array of multiple receiver coils has become a common means of reducing data acquisition time. SENSitivity Encoding (SENSE) is a popular parallel image reconstruction model that uses a complex-valued least squares estimation process to unfold aliased images. In this article, the linear mathematical framework derived in Rowe et al. [J Neurosci Meth 159 (2007) 361-369] is built upon to perform image reconstruction with subsampled data acquired from multiple receiver coils, where the SENSE model is represented as a real-valued isomorphism. A statistical analysis is performed of the various image reconstruction operators utilized in the SENSE model, with an emphasis placed on the effects of each operator on voxel means, variances and correlations. It is shown that, despite the attractiveness of models that unfold the aliased images from subsampled data, there is an artificial correlation induced between reconstructed voxels from the different folds of aliased images. As such, the mathematical framework outlined in this manuscript could be further developed to provide a means of accounting for this unavoidable correlation induced by image reconstruction operators.