Efficient Resolution Enhancement Algorithm for Compressive Sensing Magnetic Resonance Image Reconstruction

Resolution ‘scaling law’ in MRI of articular cartilage

Bone architectural information can be derived from structural indices calculated from high resolution magnetic resonance (MR) images. However, high resolution scans are time consuming and prone to motion artifacts and hence are not routinely feasible. The purpose of this study is to determine if a correlation exists between 3D structural indices calculated from high-resolution MR images and 3D co-occurrence texture features calculated from lower resolution MR images. Regression analysis indicates a strong correlation between the structural indices and the texture features. This study highlights the potential of using surrogate texture markers extracted from readily acquired clinical MR images to quantify bone architecture, circumventing the need for high resolution MR imaging.

In this paper, we present a fast iterative magnetic resonance imaging (MRI) reconstruction algorithm taking advantage of the prevailing GPGPU programming paradigm. In clinical environment, MRI reconstruction is usually performed via fast Fourier transform (FFT). However, imaging artifacts (i.e. signal loss) resulting from susceptibility-induced magnetic field inhomogeneities degrade the quality of reconstructed images. These artifacts must be addressed using accurate modeling of the physics of the system coupled with iterative reconstruction. We have developed a reconstruction algorithm with improved image quality at the expense of computation time and hence an implementation on GPUs achieving significant speedup. In this work, we extend our previous work on GPU implementation by adding several new features. First, we enable Sensitivity Encoding for Fast MRI (SENSE) reconstruction (from data acquired using a multi-receiver coil array) which can reduce the acquisition time. Besides, we have implemented a GPU-based total variation regularization in our SENSE reconstruction framework. In this paper, we describe the different optimizations employed from levels of algorithm, program code structures, and specific architecture performance tuning, featuring both our MRI reconstruction algorithm and GPU hardware specifics. Results show that the current GPU implementation produces accurate image estimates while significantly accelerating the reconstruction.

High resolution magnetic resonance (MR) imaging of the coronary artery wall in vivo has been limited by the cardiac and respiratory motion, flow artifacts as well as the relatively small size of the coronary arteries. We sought to validate in vivo black blood MR imaging of the coronary artery wall using a double inversion recovery fast spin echo MR imaging sequence with limited breath-holding and cardiac gating for suppression of motion artifacts by comparison with ex vivo MR imaging. Yorkshire albino swine (n = 6) were used in this study and coronary lesions were induced with balloon angioplasty. Four weeks after balloon injury of the coronary arteries MR imaging of the coronary artery lesions was performed. High resolution in vivo and ex vivo images of the coronary artery wall and lesions were obtained using a double inversion recovery fast spin echo sequence in a 1.5 T MR system. There was a statistically significant agreement (p < 0.0001) between measurements of vessel wall area (r = 0.87, slope = 0.87) and maximal wall thickness (r = 0.84, slope = 0.88) from in vivo and ex vivo MR images of the coronary arteries. The mean differences between in vivo and ex vivo measurements were 0.56 +/- 1.98 mm2 for vessel wall area and 0.02 +/- 0.36 mm for maximal wall thickness. Using breathholding and cardiac gating, it is possible to perform high resolution MR imaging of the coronary artery wall in vivo with good suppression of motion artifacts with a double inversion recovery fast spin echo black blood imaging sequence.

Compressed sensing (CS) algorithms exploit sparseness properties to reconstruct high spatial resolution magnetic resonance (MR) images from k-space data acquisitions significantly under sampled to reduce imaging times. CS algorithm effectiveness is frequently shown using under-sampled k-space data from NxN simulated images. These demonstration reconstructions are near perfect with quality higher than reconstructions using under-sampled NxN experimental k-space data sets. These differences are explained in terms of the interaction between the explicit transform domain sparsity requirement employed during iterative CS reconstruction and an inherent frequency domain property of the discrete Fourier transform (DFT). We report on experiments to overcome the limitations imposed by this DFT property by modifying the CS objective function to use a sparseness transform with a resolution higher that the standard transform related to the acquired NxN data matrix size. We demonstrate the relative effectiveness and limitations of standard CS and our proposed highresolution k-space extrapolation enabled (Hi-KEE) CS reconstruction on underand fully-sampled, simulated and experimental MR k-space data. (8 pages)

Objective:To optimize the scan protocol for high temporal resolution magnetic resonance (MR) imaging of the liver under single breath-holding, using compressed sensing (CS) and parallel imaging (PI) techniques in a 1.5 T MR system.Methods:31 healthy volunteers who underwent fat-suppressed gradient-echo T1 weighted imaging using a 1.5 T MR system were included. Image quality was evaluated on altering various imaging parameters in CS and PI so that the scan time was adjusted to 10 and 6 s within a single breath-holding. Normalized standard deviation (nSD = SD/mean value) and signal-to-noise ratio (SNR = mean value/SD) of liver signal intensity were measured. Visual scores for the outline of the liver and inferior right hepatic vein (IRHV) were evaluated using a 4-point scale and compared with that of the reference standard (20 s scan without CS).Results:The nSD and SNR were not significantly different when the 10 s scan with CS factor 2.0 and the 6 s scan with CS factor 2.0 and 2.5 were compared to the 20 s scan. Overall visual score (mean score of the outline of the liver and IRHV) was significantly better (p < 0.05) with the 10 s scan with CS factor 2.0 compared to the other scan protocols.Conclusion:The 10 s scan with CS factor 2.0 should be recommended for high temporal resolution MR imaging of the liver using CS and PI in a 1.5 T MR system.Advances in knowledge:This study conducts a novel MR imaging of the liver using CS and PI in a 1.5 T MR system.

Background Sagittal magnetic resonance (MR) images are typically obtained with the same spatial resolution along the entire spine, but cervical vertebrae are smaller and may be harder to assess. Purpose To investigate if high-resolution (high-res) short tau inversion recovery (STIR) and T1-weighted turbo spin echo (T1W) MR imaging (MRI) sequences are superior to standard resolution for detecting inflammatory and structural lesions in the cervical spine of patients with axial spondyloarthritis. Material and Methods Images were obtained in 36 patients. Voxel sizes at high/standard resolution were 1.99/4.33 mm3 (STIR) and 0.89/3.71 mm3 (T1W). High-resolution and standard-resolution images were scored by two readers according to the Canada-Denmark (CANDEN) MRI spine scoring system. Results Higher bone marrow edema scores were obtained at high resolution versus standard resolution (mean 2.1 vs. 1.2, P = 0.040), whereas fat lesion scores (1.8 vs. 1.5, P = 0.27) and new bone formation scores (3.5 vs. 2.8, P = 0.21) were similar. High-resolution MRI did not classify more patients as positive for bone marrow edema, fat, or new bone formation in the cervical spine compared to standard resolution. Using lateral radiographs as reference standard, sensitivity for detecting anterior corner syndesmophytes with both high-resolution and standard-resolution MRI was low (range 7–22%) and sensitivity for detecting ankylosis was low to moderate (20–55%), while specificity was high (≥96%). Conclusion High-resolution MRI allowed identification of more inflammatory lesions in the cervical spine in patients with axial spondyloarthritis when compared to standard resolution, but it did not classify more patients as positive for bone marrow edema. The slightly increased sensitivity at high-resolution MRI seemed to be too modest to have any real clinical importance.

In the last few years there has been rapid progress in the application of nuclear magnetic resonance clinical imaging as a diagnostic aid (Partain et al, 1983; Steiner, 1983; Worthington, 1983). Records kept by the operators of NMR imaging equipment indicate that about 1800 individuals have now been examined without incident within the United Kingdom. In view of this progress, an advisory group (Appendix) appointed by the National Radiological Protection Board recommends some modifications to the original advice given in 1981 (NRPB, 1981). The principal modifications are that: (1) It is no longer considered necessary to exclude volunteers with a history of epilepsy or cardiac disease. (2) It is not obligatory that a registered medical practitioner trained in resuscitation techniques be present during an exposure. (3) Higher limits are recommended for rates of change of magnetic flux density when the period of change is less than 10 ms. (4) The radiofrequency limit of 15 MHz has been removed. Theoretical considerations indicate that the interaction of an applied magnetic field with the intrinsic magnetic moments of protons and electrons within a molecule could alter its rate of chemical reaction (McLauchlan, 1981). Chemical reactions involving radical pair intermediates are expected to be sensitive to magnetic fields for this reason and a number of enzyme reactions that might be expected to be affected have been identified (A. Cass, personal communication). However, there is no experimental confirmation of these theoretical possibilities at present.

The human brain is a remarkably complex organ. Comprehension of the tissue properties of the brain and the multitude of changes in disease and ageing may lead to a greater appreciation and understanding of human biology. Combined, Magnetic Resonance Imaging (MRI) and image processing techniques are now capable of resolving very high-resolution images of the brain, including characterising discrete tissue types within small brain structures such as the hippocampus. The hippocampus is a small structure most commonly associated with its function in memory, and is comprised of subfields with distinct cellular compositions (or laminae). Computational techniques for delineating tissue types within the hippocampus incorporate sophisticated computer vision and medical imaging algorithms to better describe the range of changes occurring in these structures. In parallel, improved MR scanner hardware and sequence choices including leveraging multiple contrasts of the same tissue in vivo allows for complementary imaging information of the structure, leading to a more specific and reliable characterisation of brain tissue. Despite advances in the field, several competing algorithms and techniques are used for hippocampus subfield morphometry properties such as volume and shape.This thesis will explore computational techniques for characterising human brain tissue in vivo using multi-contrast, high resolution, ultra high-field MRI data. Specifically, this thesis will focus on characterising the tissue of the hippocampus as an exemplar for the heterogeneity of tissue classes, complex shape changes, and large, variable individual differences in anatomy between healthy and diseased/disordered subjects. This thesis will pay specific attention to characterising hippocampus tissue (in particular, image segmentation and volumetry) in a wide range of subjects and patients, which is often ignored in computational techniques for analysing MRI data.First, techniques for retrospectively ameliorating within-session motion in MRI in both healthy participants and patients with Motor Neurone Disease (MND) will be examined and discussed. It was found that incorporating a computational technique for image registration improved the reliability of image segmentation of the subfields of the hippocampus.Next, in order to better characterise the changes in anatomy over time using MRI, we constructed and evaluated a longitudinal pipeline for segmenting and examining the volume of hippocampal subfields using computational techniques in a healthy participant pool and patients from the Alzheimer's Disease Neuroimaging Initiative (ADNI). To aid in this research, the Towards Optimising MRI ChAracterisation of Tissue (TOMCAT) data set was acquired using Ultra-High Field (7-Tesla) MRI. Thirdly, to better characterise the subfields of the hippocampus and in less time than traditional computational techniques, a deep learning approach to hippocampus subfield segmentation using multi-contrast, high-resolution MRI was developed and evaluated. We found that deep learning techniques may aid in speeding the process of hippocampus subfield segmentation and deliver accurate results.Finally, the clinical applications of the previous studies were explored. We detail automated methods for extracting precise measures describing hippocampal subfield morphology and compare these features in MND patients and healthy controls. Brain tissue can be successfully characterised in vivo using high resolution and multi-contrast MRI. Developing robust and reproducible methods that are useful in both research and clinical contexts is an important step in demystifying biological processes involved in ageing and contributing to disorder and disease. This thesis adds to and improves the broader knowledge base of computational techniques for characterising brain tissue and adds novel and enhanced methodology to the field of neuroimaging.

Acquiring High Resolution (HR) Magnetic Resonance (MR) images requires the patient to remain still for long periods of time, which causes patient discomfort and increases the probability of motion induced image artifacts. A possible solution is to acquire low resolution (LR) images and to process them with the Super Resolution Generative Adversarial Network (SRGAN) to create a super-resolved version. This work applies SRGAN to MR images of the prostate and performs three experiments. The first experiment explores improving the in-plane MR image resolution by factors of 4 and 8, and shows that, while the PSNR and SSIM (Structural SIMilarity) metrics are lower than the isotropic bicubic interpolation baseline, the SRGAN is able to create images that have high edge fidelity. The second experiment explores anisotropic super-resolution via synthetic images, in that the input images to the network are anisotropically downsampled versions of HR images. This experiment demonstrates the ability of the modified SRGAN to perform anisotropic super-resolution, with quantitative image metrics that are comparable to those of the anisotropic bicubic interpolation baseline. Finally, the third experiment applies a modified version of the SRGAN to super-resolve anisotropic images obtained from the through-plane slices of the volumetric MR data. The output super-resolved images contain a significant amount of high frequency information that make them visually close to their HR counterparts. Overall, the promising results from each experiment show that super-resolution for MR images is a successful technique and that producing isotropic MR image volumes from anisotropic slices is an achievable goal.

To assess image quality and overall accuracy of 3-Tesla (3T)-sustained high-resolution magnetic resonance (MR) imaging for diagnostic preoperative workup in suspected rectal carcinoma. Twenty-three patients with suspected rectal cancer underwent unenhanced and contrast-enhanced fat-suppressed pelvic high-resolution MR imaging using a four-channel phased-array pelvic coil at 3T. Image quality, tumor stage, distance from the anorectal margin, and sphincter-saving resectability were prospectively assessed by two blinded readers. The results were correlated with clinical, surgical, and histopathologic findings. In all 23 patients, MR images were of diagnostic quality, and malignancy was correctly identified in 21 patients. The accuracy for determining sphincter-saving resectability was 100% (19/19). T stage and N stage were correctly diagnosed in 95% and 91%, respectively. MRI allowed correct identification of tumor extension and its relation to surgically relevant pelvic structures including the anorectal margin and mesorectal fascia. Transverse T2-weighted fast spin-echo images compared superiorly to all other sequences for the diagnosis of mesorectal infiltration and lymph node involvement. Moreover, transverse fat-suppressed contrast-enhanced T1-weighted images were valuable for identifying tumor infiltration, while sagittal sections were useful for the detection of longitudinal tumor extension. MR imaging with phased-array receiver coils at 3T facilitated both visualization of different pathologic conditions of the rectum and accurate determination of tumor stage in rectal carcinomas. Thus, this noninvasive diagnostic approach appeared highly suitable for the assessment of patients with suspected rectal carcinoma.

The aim of this study was to establish a correlation between a thermal measurement and a magnetic resonance imaging (MRI) signal during laser-induced interstitial thermotherapy (LITT) in liver. Pig liver was irradiated for 15 minutes with a diode laser at two different powers, 0.5 W (450 J) and 1.5 W (1,350 J). Tissue temperature was monitored every 20 seconds using thermocouples. Thermosensitive MRI sequences (T(1)-weighted Turbo-Flash) were acquired with the same irradiation parameters. Correlation between MRI signals (SI) and temperature measures was defined at two different distances from the fiber (5 and 10 mm). At 0.5 W, temperatures rose progressively up to a maximum increase of 9.5 degrees C at 5 mm and 4 degrees C at 10 mm after 15 minutes. The corresponding MRI signal decreased progressively to -27.6 SI at 5 mm and -18.5 SI at 10 mm. At 1.5 W, temperatures rose dramatically at 5 mm, reaching a plateau. The temperature elevation measured at the end of the irradiation was of 30 degrees C whereas at 10 mm it was only 14.5 degrees C. The MRI signal varied accordingly, remaining inversely proportional to temperature (-76 SI at 5 mm and -35.5 SI at 10 mm). An inversely proportional relationship was observed between MRI signal in sequential Turbo-Flash and temperature. MRI should allow to analyze heat diffusion in the liver, and thus to monitor real-time LITT treatments.

Super-resolution reconstruction of single anisotropic 3D MR images using residual convolutional neural network

Efficient techniques to noninvasively monitor stem cell transplants will accelerate the development of stem cell therapies. Magnetic resonance (MR) imaging of labeled stem cells is a noninvasive approach that can provide images with high spatial resolution. In this report, we have evaluated the ability of a commercially available, FDA-approved contrast agent to allow for the monitoring of therapeutic stem cell transplants in murine dystrophic muscle. Multipotent, muscle-derived stem cells were labeled by incubation with ferumoxide:polycation complexes. Labeled stem cells were evaluated for the ability to differentiate into mature myotubes and be detected by high resolution MR images in vitro and in vivo. Endosomal accumulation of superparamagnetic iron-oxide resulted in changes in the MR contrast T(2) and T(2)*, allowing for three dimensional, noninvasive detection of labeled cells. Relaxivity measurements on cell phantoms indicate that less than eight labeled cells could be detected by MR imaging. Furthermore, therapeutic cellular grafts transplanted into dystrophic muscle could be imaged sequentially and these noninvasive images correlated with histological images obtained by conventional microscopy. Additional studies revealed that MR imaging is applicable to tracking the migration of a small number of labeled cells following arterial delivery. MR monitoring is a highly sensitive technique that is applicable to muscle stem cell transplantation. We anticipate that it will enhance stem cell investigations by reducing the need for invasive tissue harvests and biopsies.

Cystic Adventitial Disease and High Spatial Resolution Magnetic Resonance Imaging