R. Marc Lebel’s research while affiliated with GE Healthcare and other places

Purpose To evaluate the feasibility and utility of a deep learning (DL)‐based reconstruction for improving the SNR of hyperpolarized ¹²⁹Xe lung ventilation MRI. Methods ¹²⁹Xe lung ventilation MRI data acquired from patients with asthma and/or chronic obstructive pulmonary disease (COPD) were retrospectively reconstructed with a commercial DL reconstruction pipeline at five different denoising levels. Quantitative imaging metrics of lung ventilation including ventilation defect percentage (VDP) and ventilation heterogeneity index (VHI) were compared between each set of DL‐reconstructed images and alternative denoising strategies including: filtering, total variation denoising and higher‐order singular value decomposition. Structural similarity between the denoised and original images was assessed. In a prospective study, the feasibility of using SNR gains from DL reconstruction to allow natural‐abundance xenon MRI was evaluated in healthy volunteers. Results ¹²⁹Xe ventilation image SNR was improved with DL reconstruction when compared with conventionally reconstructed images. In patients with asthma and/or COPD, DL‐reconstructed images exhibited a slight positive bias in ventilation defect percentage (1.3% at 75% denoising) and ventilation heterogeneity index (˜1.4) when compared with conventionally reconstructed images. Additionally, DL‐reconstructed images preserved structural similarity more effectively than data denoised using alternative approaches. DL reconstruction greatly improved image SNR (greater than threefold), to a level that ¹²⁹Xe ventilation imaging using natural‐abundance xenon appears feasible. Conclusion DL‐based image reconstruction significantly improves ¹²⁹Xe ventilation image SNR, preserves structural similarity, and leads to a minor bias in ventilation metrics that can be attributed to differences in the image sharpness. This tool should help facilitate cost‐effective ¹²⁹Xe ventilation imaging with natural‐abundance xenon in the future.

Deep learning (DL) reconstruction techniques to improve MR image quality are becoming commercially available with the hope that they will be applicable to multiple imaging application sites and acquisition protocols. However, before clinical implementation, these methods must be validated for specific use cases. In this work, the quality of standard-of-care (SOC) T2w and a high-spatial-resolution (HR) imaging of the breast were assessed both with and without prototype DL reconstruction. Studies were performed using data collected from phantoms, 20 retrospectively collected SOC patient exams, and 56 prospectively acquired SOC and HR patient exams. Image quality was quantitatively assessed via signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and edge sharpness. Qualitatively, all in vivo images were scored by either two or four radiologist readers using 5-point Likert scales in the following categories: artifacts, perceived sharpness, perceived SNR, and overall quality. Differences in reader scores were tested for significance. Reader preference and perception of signal intensity changes were also assessed. Application of the DL resulted in higher average SNR (1.2–2.8 times), CNR (1.0–1.8 times), and image sharpness (1.2–1.7 times). Qualitatively, the SOC acquisition with DL resulted in significantly improved image quality scores in all categories compared to non-DL images. HR acquisition with DL significantly increased SNR, sharpness, and overall quality compared to both the non-DL SOC and the non-DL HR images. The acquisition time for the HR data only required a 20% increase compared to the SOC acquisition and readers typically preferred DL images over non-DL counterparts. Overall, the DL reconstruction demonstrated improved T2w image quality in clinical breast MRI.

BOLD sensitivity to baseline perfusion and blood volume is a well-acknowledged fMRI confound. Vascular correction techniques based on cerebrovascular reactivity (CVR) might reduce variance due to baseline cerebral blood volume, however this is predicated on an invariant linear relationship between CVR and BOLD signal magnitude. Cognitive paradigms have relatively low signal, high variance and involve spatially heterogenous cortical regions; it is therefore unclear whether the BOLD response magnitude to complex paradigms can be predicted by CVR. The feasibility of predicting BOLD signal magnitude from CVR was explored in the present work across two experiments using different CVR approaches. The first utilized a large database containing breath-hold BOLD responses and 3 different cognitive tasks. The second experiment, in an independent sample, calculated CVR using the delivery of a fixed concentration of carbon dioxide and a different cognitive task. An atlas-based regression approach was implemented for both experiments to evaluate the shared variance between task-invoked BOLD responses and CVR across the cerebral cortex. Both experiments found significant relationships between CVR and task-based BOLD magnitude, with activation in the right cuneus (R ² = 0.64) and paracentral gyrus (R ² = 0.71), and the left pars opercularis (R ² = 0.67), superior frontal gyrus (R ² = 0.62) and inferior parietal cortex (R ² = 0.63) strongly predicted by CVR. The parietal regions bilaterally were highly consistent, with linear regressions significant in these regions for all four tasks. Group analyses showed that CVR correction increased BOLD sensitivity. Overall, this work suggests that BOLD signal response magnitudes to cognitive tasks are predicted by CVR across different regions of the cerebral cortex, providing support for the use of correction based on baseline vascular physiology.

Objectives: To compare interobserver agreement and image quality of 3D T2-weighted fast spin echo (T2w-FSE) L-spine MRI images processed with a deep learning reconstruction (DLRecon) against standard-of-care (SOC) reconstruction, as well as against 2D T2w-FSE images. The hypothesis was that DLRecon 3D T2w-FSE would afford improved image quality and similar interobserver agreement compared to both SOC 3D and 2D T2w-FSE. Methods: Under IRB approval, patients who underwent routine 3-T lumbar spine (L-spine) MRI from August 17 to September 17, 2020, with both isotropic 3D and 2D T2w-FSE sequences, were retrospectively included. A DLRecon algorithm, with denoising and sharpening properties was applied to SOC 3D k-space to generate 3D DLRecon images. Four musculoskeletal radiologists blinded to reconstruction status evaluated randomized images for motion artifact, image quality, central/foraminal stenosis, disc degeneration, annular fissure, disc herniation, and presence of facet joint cysts. Inter-rater agreement for each graded variable was evaluated using Conger's kappa (κ). Results: Thirty-five patients (mean age 58 ± 19, 26 female) were evaluated. 3D DLRecon demonstrated statistically significant higher median image quality score (2.0/2) when compared to SOC 3D (1.0/2, p < 0.001), 2D axial (1.0/2, p < 0.001), and 2D sagittal sequences (1.0/2, p value < 0.001). κ ranges (and 95% CI) for foraminal stenosis were 0.55-0.76 (0.32-0.86) for 3D DLRecon, 0.56-0.73 (0.35-0.84) for SOC 3D, and 0.58-0.71 (0.33-0.84) for 2D. Mean κ (and 95% CI) for central stenosis at L4-5 were 0.98 (0.96-0.99), 0.97 (0.95-0.99), and 0.98 (0.96-0.99) for 3D DLRecon, 3D SOC and 2D, respectively. Conclusions: DLRecon 3D T2w-FSE L-spine MRI demonstrated higher image quality and similar interobserver agreement for graded variables of interest when compared to 3D SOC and 2D imaging. Key points: • 3D DLRecon T2w-FSE isotropic lumbar spine MRI provides improved image quality when compared to 2D MRI, with similar interobserver agreement for clinical evaluation of pathology. • 3D DLRecon images demonstrated better image quality score (2.0/2) when compared to standard-of-care (SOC) 3D (1.0/2), p value < 0.001; 2D axial (1.0/2), p value < 0.001; and 2D sagittal sequences (1.0/2), p value < 0.001. • Interobserver agreement for major variables of interest was similar among all sequences and reconstruction types. For foraminal stenosis, κ ranged from 0.55 to 0.76 (95% CI 0.32-0.86) for 3D DLRecon, 0.56-0.73 (95% CI 0.35-0.84) for standard-of-care (SOC) 3D, and 0.58-0.71 (95% CI 0.33-0.84) for 2D.

Purpose: To evaluate two settings (noise reduction of 50% or 75%) of a deep learning (DL) reconstruction model relative to each other and to conventional MR image reconstructions on clinical orthopedic MRI datasets. Materials and methods: This retrospective study included 54 patients who underwent two-dimensional fast spin-echo MRI for hip (n = 22; mean age, 44 years ± 13 [standard deviation]; nine men) or shoulder (n = 32; mean age, 56 years ± 17; 17 men) conditions between March 2019 and June 2020. MR images were reconstructed with conventional methods and the vendor-provided and commercially available DL model applied with 50% and 75% noise reduction settings (DL 50 and DL 75, respectively). Quantitative analytics, including relative anatomic edge sharpness, relative signal-to-noise ratio (rSNR), and relative contrast-to-noise ratio (rCNR) were computed for each dataset. In addition, the image sets were randomized, blinded, and presented to three board-certified musculoskeletal radiologists for ranking based on overall image quality and diagnostic confidence. Statistical analysis was performed with a nonparametric hypothesis comparing derived quantitative metrics from each reconstruction approach. In addition, inter- and intrarater agreement analysis was performed on the radiologists' rankings. Results: Both denoising settings of the DL reconstruction showed improved edge sharpness, rSNR, and rCNR relative to the conventional reconstructions. The reader rankings demonstrated strong agreement, with both DL reconstructions outperforming the conventional approach (Gwet agreement coefficient = 0.98). However, there was lower agreement between the readers on which DL reconstruction denoising setting produced higher-quality images (Gwet agreement coefficient = 0.31 for DL 50 and 0.35 for DL 75). Conclusion: The vendor-provided DL MRI reconstruction showed higher edge sharpness, rSNR, and rCNR in comparison with conventional methods; however, optimal levels of denoising may need to be further assessed.Keywords: MRI Reconstruction Method, Deep Learning, Image Analysis, Signal-to-Noise Ratio, MR-Imaging, Neural Networks, Hip, Shoulder, Physics, Observer Performance, Technology Assessment Supplemental material is available for this article. © RSNA, 2021.

Purpose To develop a deep‐learning model that leverages the spatial and temporal information from dynamic contrast‐enhanced magnetic resonance (DCE MR) brain imaging in order to automatically estimate a vascular function (VF) for quantitative pharmacokinetic (PK) modeling. Methods Patients with glioblastoma multiforme were scanned post‐resection approximately every 2 months using a high spatial and temporal resolution DCE MR imaging sequence (≈5 s and ≈2 cm³). A region over the transverse sinus was manually drawn in the dynamic T1‐weighted images to provide a ground truth VF. The manual regions and their resulting VF curves were used to train a deep‐learning model based on a 3D U‐net architecture. The model concurrently utilized the spatial and temporal information in DCE MR images to predict the VF. In order to analyze the contribution of the spatial and temporal terms, different weighted combinations were examined. The manual and deep‐learning predicted regions and VF curves were compared. Results Forty‐three patients were enrolled in this study and 155 DCE MR scans were processed. The 3D U‐net was trained using a loss function that combined the spatial and temporal information with different weightings. The best VF curves were obtained when both spatial and temporal information were considered. The predicted VF curve was similar to the manual ground truth VF curves. Conclusion The use of spatial and temporal information improved VF curve prediction relative to when only the spatial information is used. The method generalized well for unseen data and can be used to automatically estimate a VF curve suitable for quantitative PK modeling. This method allows for a more efficient clinical pipeline and may improve automation of permeability mapping.

Objective To assess a new deep learning-based MR reconstruction method, “DLRecon,” for clinical evaluation of peripheral nerves. Methods Sixty peripheral nerves were prospectively evaluated in 29 patients (mean age: 49 ± 16 years, 17 female) undergoing standard-of-care (SOC) MR neurography for clinically suspected neuropathy. SOC-MRIs and DLRecon-MRIs were obtained through conventional and DLRecon reconstruction methods, respectively. Two radiologists randomly evaluated blinded images for outer epineurium conspicuity, fascicular architecture visualization, pulsation artifact, ghosting artifact, and bulk motion. Results DLRecon-MRIs were likely to score better than SOC-MRIs for outer epineurium conspicuity (OR = 1.9, p = 0.007) and visualization of fascicular architecture (OR = 1.8, p < 0.001) and were likely to score worse for ghosting (OR = 2.8, p = 0.004) and pulsation artifacts (OR = 1.6, p = 0.004). There was substantial to almost-perfect inter-reconstruction method agreement (AC = 0.73–1.00) and fair to almost-perfect interrater agreement (AC = 0.34–0.86) for all features evaluated. DLRecon-MRI had improved interrater agreement for outer epineurium conspicuity (AC = 0.71, substantial agreement) compared to SOC-MRIs (AC = 0.34, fair agreement). In >80% of images, the radiologist correctly identified an image as SOC- or DLRecon-MRI. Discussion Outer epineurium and fascicular architecture conspicuity, two key morphological features critical to evaluating a nerve injury, were improved in DLRecon-MRIs compared to SOC-MRIs. Although pulsation and ghosting artifacts increased in DLRecon images, image interpretation was unaffected.

Background Deep learning–based methods have been used to denoise magnetic resonance imaging. Purpose The purpose of this study was to evaluate a deep learning reconstruction (DL Recon) in cardiovascular black-blood T2-weighted images and compare with intensity filtered images. Material and Methods Forty-five DL Recon images were compared with intensity filtered and the original images. For quantitative image analysis, the signal to noise ratio (SNR) of the septum, contrast ratio (CR) of the septum to lumen, and sharpness of the endocardial border were calculated in each image. For qualitative image quality assessment, a 4-point subjective scale was assigned to each image (1 = poor, 2 = fair, 3 = good, 4 = excellent). Results The SNR and CR were significantly higher in the DL Recon images than in the intensity filtered and the original images ( p < .05 in each). Sharpness of the endocardial border was significantly higher in the DL Recon and intensity filtered images than in the original images ( p < .05 in each). The image quality of the DL Recon images was significantly better than that of intensity filtered and original images ( p < .001 in each). Conclusions DL Recon reduced image noise while improving image contrast and sharpness in the cardiovascular black-blood T2-weight sequence.

Background Advances in sub-Nyquist–sampled dynamic contrast-enhanced (DCE) MRI enable monitoring of brain tumors with millimeter resolution and whole-brain coverage. Such undersampled quantitative methods need careful characterization regarding achievable test-retest reproducibility. Purpose To demonstrate a fully automated high-resolution whole-brain DCE MRI pipeline with 30-fold sparse undersampling and estimate its reproducibility on the basis of reference regions of stable tissue types during multiple posttreatment time points by using longitudinal clinical images of high-grade glioma. Materials and Methods Two methods for sub-Nyquist–sampled DCE MRI were extended with automatic estimation of vascular input functions. Continuously acquired three-dimensional k-space data with ramped-up flip angles were partitioned to yield high-resolution, whole-brain tracer kinetic parameter maps with matched precontrast-agent T1 and M0 maps. Reproducibility was estimated in a retrospective study in participants with high-grade glioma, who underwent three consecutive standard-of-care examinations between December 2016 and April 2019. Coefficients of variation and reproducibility coefficients were reported for histogram statistics of the tracer kinetic parameters plasma volume fraction and volume transfer constant (Ktrans) on five healthy tissue types. Results The images from 13 participants (mean age ± standard deviation, 61 years ± 10; nine women) with high-grade glioma were evaluated. In healthy tissues, the protocol achieved a coefficient of variation less than 57% for median Ktrans, if Ktrans was estimated consecutively. The maximum reproducibility coefficient for median Ktrans was estimated to be at 0.06 min–1 for large or low-enhancing tissues and to be as high as 0.48 min–1 in smaller or strongly enhancing tissues. Conclusion A fully automated, sparsely sampled DCE MRI reconstruction with patient-specific vascular input function offered high spatial and temporal resolution and whole-brain coverage; in healthy tissues, the protocol estimated median volume transfer constant with maximum reproducibility coefficient of 0.06 min–1 in large, low-enhancing tissue regions and maximum reproducibility coefficient of less than 0.48 min–1 in smaller or more strongly enhancing tissue regions. Published under a CC BY 4.0 license. Online supplemental material is available for this article. See also the editorial by Lenkinski in this issue.

Purpose To develop and evaluate an efficient precontrast T1 mapping technique suitable for quantitative high‐resolution whole‐brain dynamic contrast‐enhanced–magnetic resonance imaging (DCE‐MRI). Methods Variable flip angle (VFA) T1 mapping was considered that provides 1 × 1 × 2 mm³ resolution to match a recent high‐resolution whole‐brain DCE‐MRI protocol. Seven FAs were logarithmically spaced from 1.5° to 15°. T1 and M0 maps were estimated using model‐based reconstruction. This approach was evaluated using an anatomically realistic brain tumor digital reference object (DRO) with noise‐mimicking 3T neuroimaging and fully sampled data acquired from one healthy volunteer. Methods were also applied on fourfold prospectively undersampled VFA data from 13 patients with high‐grade gliomas. Results T1‐mapping precision decreased with undersampling factor R, althoughwhereas bias remained small before a critical R. In the noiseless DRO, T1 bias was <25 ms in white matter (WM) and <11 ms in brain tumor (BT). T1 standard deviation (SD) was <119.5 ms in WM (coefficient of variation [COV] ~11.0%) and <253.2 ms in BT (COV ~12.7%). In the noisy DRO, T1 bias was <50 ms in WM and <30 ms in BT. For R ≤ 10, T1 SD was <107.1 ms in WM (COV ~9.9%) and <240.9 ms in BT (COV ~12.1%). In the healthy subject, T1 bias was <30 ms for R ≤ 16. At R = 4, T1 SD was 171.4 ms (COV ~13.0%). In the prospective brain tumor study, T1 values were consistent with literature values in WM and BT. Conclusion High‐resolution whole‐brain VFA T1 mapping is feasible with sparse sampling, supporting its use for quantitative DCE‐MRI.