Skip to main content
eScholarship
Open Access Publications from the University of California

Scholarly Works (40 results)

  • Thesis
  • Peer Reviewed
UCLA Electronic Theses and Dissertations (2015)

Cardiovascular diseases are the leading cause of death globally and atherosclerosis is the primary underlying condition. Magnetic resonance imaging (MRI) is a promising diagnostic tool for evaluating atherosclerosis because of its excellent tissue contrast non-invasiveness. However many vascular MRI techniques have not been utilized clinically due to practical issues related to imaging speed, workflow, contrast media, and image quality. The primary focus of the work in this dissertation is to address some of these limitations and improve the capability of vascular MRI to accurately detect and characterize atherosclerotic lesions. Specifically, there are four areas of improvements of vascular MRI:

First, an online respiratory self-gating method (ADIOS) was developed for noncontrast magnetic resonance angiography of renal arteries. Some drawbacks of the conventional navigator gating method, including complicated planning and saturation artifacts were eliminated. Testing on healthy subjects confirmed that the proposed method was able to provide high-quality visualization of renal arteries with no navigator-induced artifacts, simplified setup, and shorter scan time.

Second, a diffusion-prepared turbo-spin-echo technique (DP-TSE) was developed to enable 3D high resolution diffusion weighted imaging for carotid plaque characterization without the need of contrast agent. The proposed method showed significantly higher vessel wall visibility, less distortion, and less partial volume effect. Apparent diffusion coefficient demonstrated the potential to differentiate lipid core from fibrous plaque tissue.

Third, a black-blood imaging technique (DANTE-SPACE) was developed to improve flow signal suppression in vessel wall imaging. A combined carotid and intracranial arterial wall imaging protocol was designed and tested. DANTE-SPACE demonstrated significantly improved arterial and venous blood suppression compared with conventional methods. Preliminary clinical study showed that the proposed method had the potential to detect vessel wall dissection and accurately differentiate plaque components.

Fourth, an imaging method for coronary plaque characterization was developed to allow simultaneous dark-blood T1-weighted imaging and bright-blood imaging. An image-based affine motion correction algorithm was designed to allow 100% respiratory gating efficiency. Results from healthy subjects demonstrated the feasibility of whole-heart coronary plaque imaging with isotropic high resolution. Preliminary results from patients with coronary artery disease showed the potential of this technique to detect intra-plaque hemorrhage and inflammation in the coronary wall.

    Cover page: Improvements on Vascular Magnetic Resonance Imaging
    • Article
    • Peer Reviewed
    UCLA Previously Published Works (2022)
    Non-Cartesian sampling with subspace-constrained image reconstruction is a popular approach to dynamic MRI, but slow iterative reconstruction limits its clinical application. Data-consistent (DC) deep learning can accelerate reconstruction with good image quality, but has not been formulated for non-Cartesian subspace imaging. In this study, we propose a DC non-Cartesian deep subspace learning framework for fast, accurate dynamic MR image reconstruction. Four novel DC formulations are developed and evaluated: two gradient decent approaches, a directly solved approach, and a conjugate gradient approach. We applied a U-Net model with and without DC layers to reconstruct T1-weighted images for cardiac MR Multitasking (an advanced multidimensional imaging method), comparing our results to the iteratively reconstructed reference. Experimental results show that the proposed framework significantly improves reconstruction accuracy over the U-Net model without DC, while significantly accelerating the reconstruction over conventional iterative reconstruction.
      Cover page: Data-Consistent Non-Cartesian Deep Subspace Learning for Efficient Dynamic MR Image Reconstruction
      • Article
      • Peer Reviewed
      UCLA Previously Published Works (2016)

      Purpose

      The purpose of this study was to develop a three-dimensional black blood imaging method for simultaneously evaluating the carotid and intracranial arterial vessel walls with high spatial resolution and excellent blood suppression with and without contrast enhancement.

      Methods

      The delay alternating with nutation for tailored excitation (DANTE) preparation module was incorporated into three-dimensional variable flip angle turbo spin echo (SPACE) sequence to improve blood signal suppression. Simulations and phantom studies were performed to quantify image contrast variations induced by DANTE. DANTE-SPACE, SPACE, and two-dimensional turbo spin echo were compared for apparent signal-to-noise ratio, contrast-to-noise ratio, and morphometric measurements in 14 healthy subjects. Preliminary clinical validation was performed in six symptomatic patients.

      Results

      Apparent residual luminal blood was observed in five (pre-contrast) and nine (post-contrast) subjects with SPACE and only two (post-contrast) subjects with DANTE-SPACE. DANTE-SPACE showed 31% (pre-contrast) and 100% (post-contrast) improvement in wall-to-blood contrast-to-noise ratio over SPACE. Vessel wall area measured from SPACE was significantly larger than that from DANTE-SPACE due to possible residual blood signal contamination. DANTE-SPACE showed the potential to detect vessel wall dissection and identify plaque components in patients.

      Conclusion

      DANTE-SPACE significantly improved arterial and venous blood suppression compared with SPACE. Simultaneous high-resolution carotid and intracranial vessel wall imaging to potentially identify plaque components was feasible with a scan time under 6 min. Magn Reson Med 75:2286-2294, 2016. © 2015 Wiley Periodicals, Inc.
        Cover page: Improved black‐blood imaging using DANTE‐SPACE for simultaneous carotid and intracranial vessel wall evaluation
        • Article
        • Peer Reviewed
        UCLA Previously Published Works (2016)

        Purpose

        To improve the coronary visualization quality of four-dimensional (4D) coronary MR angiography (MRA) through cardiac motion correction and iterative reconstruction.

        Methods

        A contrast-enhanced, spoiled gradient echo sequence with 3D radial trajectory and self-gating was used for 4D coronary MRA data acquisition at 3 Tesla. A whole-heart 16-phase cine series was reconstructed with respiratory motion correction. Nonrigid registration was performed between the identified quiescent phases and a reference. The motion information of all included phases was then used along with the corresponding k-space data to iteratively reconstruct the final image. Healthy volunteer studies (N = 13) were conducted to compare the proposed method with the conventional strategy, which accepts data from a single, contiguous window out of the original 16-phase data. Apparent signal-to-noise ratio (aSNR) and coronary sharpness were used as the image quality metrics.

        Results

        The proposed method significantly improved aSNR (11.89 ± 3.76 to 13.97 ± 5.21; P = 0.005) and scan efficiency (18.8% ± 6.0% to 40.9% ± 9.7%; P < 0.001), compared with the conventional strategy. Sharpness of left main (P = 0.002), proximal (P = 0.04), and middle (P = 0.02) right coronary artery, and proximal left anterior descending (P = 0.04) was also significantly improved.

        Conclusion

        The proposed cardiac motion-corrected reconstruction significantly improved the achievable quality of coronary visualization from 4D coronary MRA. Magn Reson Med 76:1345-1353, 2016. © 2016 International Society for Magnetic Resonance in Medicine.
          Cover page: High efficiency coronary MR angiography with nonrigid cardiac motion correction
          • Article
          • Peer Reviewed
          UCLA Previously Published Works (2018)
          Magnetic resonance image (MRI) in high spatial resolution provides detailed anatomical information and is often necessary for accurate quantitative analysis. However, high spatial resolution typically comes at the expense of longer scan time, less spatial coverage, and lower signal to noise ratio (SNR). Single Image Super-Resolution (SISR), a technique aimed to restore high-resolution (HR) details from one single low-resolution (LR) input image, has been improved dramatically by recent breakthroughs in deep learning. In this paper, we introduce a new neural network architecture, 3D Densely Connected Super-Resolution Networks (DCSRN) to restore HR features of structural brain MR images. Through experiments on a dataset with 1,113 subjects, we demonstrate that our network outperforms bicubic interpolation as well as other deep learning methods in restoring 4× resolution-reduced images.
            • Article
            • Peer Reviewed
            UCLA Previously Published Works (2022)

            Purpose

            To address head motion in brain MRI with a novel motion-resolved imaging framework, with application to motion-robust quantitative multiparametric mapping.

            Methods

            MR multitasking conceptualizes the underlying multiparametric image in the presence of motion as a multidimensional low-rank tensor. By incorporating a motion-state dimension into the parameter dimensions and introducing unsupervised motion-state binning and outlier motion reweighting mechanisms, the brain motion can be readily resolved for motion-robust quantitative imaging. A numerical motion phantom was used to simulate different discrete and periodic motion patterns under various translational and rotational scenarios, as well as investigate the sensitivity to exceptionally small and large displacements. In vivo brain MRI was performed to also evaluate different real discrete and periodic motion patterns. The effectiveness of motion-resolved imaging for simultaneous T1 /T2 /T1ρ mapping in the brain was investigated.

            Results

            For all 14 simulation scenarios of small, intermediate, and large motion displacements, the motion-resolved approach produced T1 /T2 /T1ρ maps with less absolute difference errors against the ground truth, lower RMSE, and higher structural similarity index measure of T1 /T2 /T1ρ measurements compared to motion removal, and no motion handling. For in vivo experiments, the motion-resolved approach produced T1 /T2 /T1ρ maps with the best image quality free from motion artifacts under random discrete motion, tremor, periodic shaking, and nodding patterns compared to motion removal and no motion handling. The proposed method also yielded T1 /T2 /T1ρ measurement distributions closest to the motion-free reference, with minimal measurement bias and variance.

            Conclusion

            Motion-resolved quantitative brain imaging is achieved with multitasking, which is generalizable to various head motion patterns without explicit need for registration-based motion correction.
              Cover page: Motion‐robust quantitative multiparametric brain MRI with motion‐resolved MR multitasking
              • Article
              • Peer Reviewed
              UCLA Previously Published Works (2016)

              Purpose

              To optimize a diffusion-prepared balanced steady-state free precession cardiac MRI (CMR) technique to perform diffusion-tensor CMR (DT-CMR) in humans on a 3 Tesla clinical scanner METHODS: A previously developed second order motion compensated (M2) diffusion-preparation scheme was significantly shortened (40%) yielding sufficient signal-to-noise ratio for DT-CMR imaging. In 20 healthy volunteers and 3 heart failure (HF) patients, DT-CMR was performed comparing no motion compensation (M0), first order motion compensation (M1), and the optimized M2. Mean diffusivity (MD), fractional anisotropy (FA), helix angle (HA), and HA transmural slope (HATS) were calculated. Reproducibility and success rate (SR) were investigated.

              Results

              M2-derived left ventricular (LV) MD, FA, and HATS (1.4 ± 0.2 μm2 /ms, 0.28 ± 0.06, -1.0 ± 0.2 °/%trans) were significantly (P < 0.001) less than M1 (1.8 ± 0.3 μm2 /ms, 0.46 ± 0.14, -0.1 ± 0.3 °/%trans) and M0 (4.8 ± 1.0 μm2 /ms, 0.70 ± 0.14, 0.1 ± 0.3 °/%trans) indicating less motion corruption and yielding values more consistent with previous literature. M2-derived DT-CMR parameters had higher reproducible (ICC > 0.85) and SR (82%) than M1 (ICC = 0.20-0.85; SR = 37%) and M0 (ICC = 0.20-0.30; SR = 11%). M2 DT-CMR was able to yield HA maps with smooth transmural transition from endocardium to epicardium.

              Conclusion

              The proposed M2 DT-CMR reproducibly yielded bulk motion robust estimations of mean LV MD, FA, HA, and HATS on a 3T clinical scanner. Magn Reson Med 76:1354-1363, 2016. © 2016 International Society for Magnetic Resonance in Medicine.
                Cover page: In vivo diffusion‐tensor MRI of the human heart on a 3 tesla clinical scanner: An optimized second order (M2) motion compensated diffusion‐preparation approach
                • Article
                UCLA Previously Published Works (2020)
                High-resolution (HR) magnetic resonance imaging (MRI) provides detailed anatomical information that is critical for diagnosis in the clinical application. However, HR MRI typically comes at the cost of long scan time, small spatial coverage, and low signal-to-noise ratio (SNR). Recent studies showed that with a deep convolutional neural network (CNN), HR generic images could be recovered from low-resolution (LR) inputs via single image super-resolution (SISR) approaches. Additionally, previous works have shown that a deep 3D CNN can generate high-quality SR MRIs by using learned image priors. However, 3D CNN with deep structures, have a large number of parameters and are computationally expensive. In this paper, we propose a novel 3D CNN architecture, namely a multi-level densely connected super-resolution network (mDCSRN), which is light-weight, fast and accurate. We also show that with the generative adversarial network (GAN)-guided training, the mDCSRN-GAN provides appealing sharp SR images with rich texture details that are highly comparable with the referenced HR images. Our results from experiments on a large public dataset with 1,113 subjects showed that this new architecture outperformed other popular deep learning methods in recovering 4x resolution-downgraded images in both quality and speed.
                  Cover page: MRI Super-Resolution with GAN and 3D Multi-Level DenseNet: Smaller, Faster, and Better
                  • Article
                  • Peer Reviewed
                  UCLA Previously Published Works (2018)
                  High-resolution (HR) magnetic resonance images (MRI) provide detailed anatomical information important for clinical application and quantitative image analysis. However, HR MRI conventionally comes at the cost of longer scan time, smaller spatial coverage, and lower signal-to-noise ratio (SNR). Recent studies have shown that single image super-resolution (SISR), a technique to recover HR details from one single low-resolution (LR) input image, could provide high quality image details with the help of advanced deep convolutional neural networks (CNN). However, deep neural networks consume memory heavily and run slowly, especially in 3D settings. In this paper, we propose a novel 3D neural network design, namely a multi-level densely connected super-resolution network (mDCSRN) with generative adversarial network (GAN)–guided training. The mDCSRN trains and inferences quickly, and the GAN promotes realistic output hardly distinguishable from original HR images. Our results from experiments on a dataset with 1,113 subjects shows that our new architecture outperforms other popular deep learning methods in recovering 4x resolution-downgraded images and runs 6x faster.
                    • Article
                    • Peer Reviewed
                    UCLA Previously Published Works (2018)
                    Quantitative cardiovascular magnetic resonance (CMR) imaging can be used to characterize fibrosis, oedema, ischaemia, inflammation and other disease conditions. However, the need to reduce artefacts arising from body motion through a combination of electrocardiography (ECG) control, respiration control, and contrast-weighting selection makes CMR exams lengthy. Here, we show that physiological motions and other dynamic processes can be conceptualized as multiple time dimensions that can be resolved via low-rank tensor imaging, allowing for motion-resolved quantitative imaging with up to four time dimensions. This continuous-acquisition approach, which we name cardiovascular MR multitasking, captures - rather than avoids - motion, relaxation and other dynamics to efficiently perform quantitative CMR without the use of ECG triggering or breath holds. We demonstrate that CMR multitasking allows for T1 mapping, T1-T2 mapping and time-resolved T1 mapping of myocardial perfusion without ECG information and/or in free-breathing conditions. CMR multitasking may provide a foundation for the development of setup-free CMR imaging for the quantitative evaluation of cardiovascular health.
                      Cover page: Magnetic resonance multitasking for motion-resolved quantitative cardiovascular imaging
                      Top