Gibbs Artifacts Research Articles

NIMG-33. HIGH-RESOLUTION SODIUM MRI OF HUMAN GLIOMAS AT 3T USING PHYSICS-BASED GENERATIVE AI

Abstract BACKGROUND Sodium neuroimaging is a promising technique for diagnosing and monitoring brain tumors, providing insights into microenvironment and metabolism. However, at 3T, it is limited by low signal-to-noise ratio and resolution, resulting in long acquisition times and low-quality images. To address these limitations, we developed a physics-informed generative adversarial network (GAN) approach for high-resolution sodium neuroimaging of brain tumors at 3T. METHODS 5,078 anatomical sequences from 1,330 brain tumor patients undergoing routine proton MRI were used to create a synthetic dataset with physics-based simulated artifacts, including B0 magnetic susceptibility artifacts, chemical shift artifacts, Nyquist aliasing artifacts, Gibbs artifacts, and low-resolution images, with Rician noise added during training to augment the dataset. By using multiple proton MRI contrasts, we hypothesized that the GAN model would learn to correct artifacts while preserving the inherent contrast information. We employed a transfer learning approach with a modified Pix2PixGAN architecture and an Attention-R2UNet generator. Twenty glioma patients consented to participate in this IRB-approved prospective study. Sodium and proton MRIs were acquired in the same sessions using a 3T Siemens scanner using a dual-tuned head coil, with a 3D spoiled gradient-echo sequence optimized for short TE. Sodium-proton exchanger (NHE1) expression was tested with immunohistochemistry from image-guided surgical biopsies. RESULTS High-resolution synthetic-sodium MR images were generated using the GAN model. The proposed method successfully reconstructed high-resolution synthetic-sodium MR images with improved SNR compared to traditional sodium images. Synthetic and native measurements were strongly correlated (R2=0.8539, P<0.0001). A significant correlation was seen between synthetic-sodium MR measurements and relative NHE1 tissue expression (ρ=0.5817, P=0.0036). CONCLUSION Results suggest that high-resolution synthetic-sodium MRI retains much of the inherent information from native sodium MR while accurately representing NHE1 expression. This approach could make high-resolution sodium neuroimaging at 3T feasible in the clinical environment, potentially improving diagnosis, monitoring, and treatment of brain tumors.

Optimization of penalization function in Bayesian penalized likelihood reconstruction algorithm for [18F]flutemetamol amyloid PET images.

Point-spread-function (PSF) correction is not recommended for amyloid PET images due to Gibbs artifacts. Q.Clear™, a Bayesian Penalized Likelihood (BPL) reconstruction method without incorporating PSF correction reduces these artifacts but degrades image contrast by our previous findings. The present study aimed to recover lost contrast by optimizing reconstruction parameters in time-of-flight (TOF) BPL reconstruction of amyloid PET images without PSF correction. We selected candidate conditions based on a phantom study and then determined which were optimal in a clinical study. Phantom images were reconstructed under conditions of 1‒9 iterations, β 300-1000 and γ factors from 2 to 10 in TOF-BPL without PSF correction. We evaluated the %contrast and the coefficients of variation (CV, %). Standardized uptake value ratios (SUVr) and Centiloid scales (CL) were calculated from PET images acquired from 71 participants after an [18F]flutemetamol injection. Both %contrast and CV were independent of iterations, whereas a trade-off was found between γ factors and β. We selected a γ factors of 5 without PSF correction (iterations, 1; β, 500) and of 10 without PSF correction (iterations, 1; β, 800) as candidates for clinical investigation. The SUVr and CL remained stable across various conditions, and CL scales effectively discriminated amyloid PET using measured values. The optimal reconstruction parameters of TOF-BPL for [18F]flutemetamol PET images were γ factor 10, iterations 1 and β 800, without PSF correction.

Phantom and clinical evaluation of the Bayesian penalised likelihood reconstruction algorithm Q.Clear without PSF correction in amyloid PET images

PurposeBayesian penalised likelihood (BPL) reconstruction, which incorporates point-spread-function (PSF) correction, provides higher signal-to-noise ratios and more accurate quantitation than conventional ordered subset expectation maximization (OSEM) reconstruction. However, applying PSF correction to brain PET imaging is controversial due to Gibbs artefacts that manifest as unpredicted cortical uptake enhancement. The present study aimed to validate whether BPL without PSF would be useful for amyloid PET imaging.MethodsImages were acquired from Hoffman 3D brain and cylindrical phantoms for phantom study and 71 patients administered with [18F]flutemetamol in clinical study using a Discovery MI. All images were reconstructed using OSEM, BPL with PSF correction, and BPL without PSF correction. Count profile, %contrast, recovery coefficients (RCs), and image noise were calculated from the images acquired from the phantoms. Amyloid β deposition in patients was visually assessed by two physicians and quantified based on the standardised uptake value ratio (SUVR).ResultsThe overestimated radioactivity in profile curves was eliminated using BPL without PSF correction. The %contrast and image noise decreased with increasing β values in phantom images. Image quality and RCs were better using BPL with, than without PSF correction or OSEM. An optimal β value of 600 was determined for BPL without PSF correction. Visual evaluation almost agreed perfectly (κ = 0.91–0.97), without depending on reconstruction methods. Composite SUVRs did not significantly differ between reconstruction methods.ConclusionGibbs artefacts disappeared from phantom images using the BPL without PSF correction. Visual and quantitative evaluation of [18F]flutemetamol imaging was independent of the reconstruction method. The BPL without PSF correction could be the standard reconstruction method for amyloid PET imaging, despite being qualitatively inferior to BPL with PSF correction for [18F]flutemetamol amyloid PET imaging.

Intra-patient and inter-observer image quality analysis in liver MRI study with gadoxetic acid using two different multi-arterial phase techniques.

To evaluate intra-patient and interobserver agreement in patients who underwent liver MRI with gadoxetic acid using two different multi-arterial phase (AP) techniques. A total of 154 prospectively enrolled patients underwent clinical gadoxetic acid-enhanced liver MRI twice within 12 months, using two different multi-arterial algorithms: CAIPIRINHA-VIBE and TWIST-VIBE. For every patient, breath-holding time, body mass index, sex, age were recorded. The phase without contrast media and the APs were independently evaluated by two radiologists who quantified Gibbs artefacts, noise, respiratory motion artefacts, and general image quality. Presence or absence of Gibbs artefacts and noise was compared by the McNemar's test. Respiratory motion artefacts and image quality scores were compared using Wilcoxon signed rank test. Interobserver agreement was assessed by Cohen kappa statistics. Compared with TWIST-VIBE, CAIPIRINHA-VIBE images had better scores for every parameter except higher noise score. Triple APs were always acquired with TWIST-VIBE but failed in 37% using CAIPIRINHA-VIBE: 11% have only one AP, 26% have two. Breath-holding time was the only parameter that influenced the success of multi-arterial techniques. TWIST-VIBE images had worst score for Gibbs and respiratory motion artefacts but lower noise score. CAIPIRINHA-VIBE images were always diagnostic, but with a failure of triple-AP in 37%. TWIST-VIBE was successful in obtaining three APs in all patients. Breath-holding time is the only parameter which can influence the preliminary choice between CAIPIRINHA-VIBE and TWIST-VIBE algorithm. If the patient is expected to perform good breath-holds, TWIST-VIBE is preferable; otherwise, CAIPIRINHA-VIBE is more appropriate.

An adaptive Ritz formulation for progressive damage modelling in variable angle tow composite plates

In this work, an adaptive Ritz model for the analysis of variable angle tow composite plates featuring damage initiation and evolution under progressive loading is proposed, developed, implemented and tested. The plate kinematics is represented employing a first-order shear deformation theory, while the plate equilibrium equations at a given load step are obtained by minimizing the structure potential energy. The constitutive behaviour is modelled within the framework of continuum damage mechanics. In particular the initiation and evolution of damage, up to failure, are tracked by defining irreversible damage indices related to both fibres and matrix, both in tensile or compression loading. The discrete equations are then obtained by assuming a polynomial Ritz approximation of the primary kinematic variables in the energy minimization. Preliminary tests show how the application of the method as a single-domain approach induces the emergence of problematic spurious effects, related to Gibbs artefacts due to the inability of the selected polynomial basis to represent damage localization. An adaptive multi-domain technique is thus proposed to circumvent such issues, which has been successfully validated by benchmark tests. Eventually, original results about variable angle tow plates featuring damage evolution under progressive loading are presented.

Gibbs Artifacts Removal with Nonlinearity.

Gibbs artifacts, appearing as oscillations or ringing around sharp edges or boundaries, are frequently encountered in image processing. They arise when the image's frequency components are adjusted, such as in image deblurring and sharpening. Linear methods are ineffective in reducing Gibbs artifacts; nonlinear methods may be more effective. One such nonlinear method is the use of neural networks. This paper applies a simple convolutional neural network (CNN) to an image sharpening task and observes the effects of Gibbs artifacts. This network has only one convolutional layer, which consists of four channels. The well-known rectified linear unit (ReLU) is used as the nonlinear activation function. For simple one-dimensional (1D) and two-dimensional (2D), unrealistic case studies, the Gibbs artifacts are completely removed. The reason why the artifacts can be removed is explained. This simple case study illustrates the power of nonlinear functions and the use of multiple channels. In fact, this task can be achieved without using a neural network.

Automatic detection of Gibbs artefact in MR images with transfer learning approach.

Quality control of magnetic resonance imaging includes image validation, which covers also artefact detection. The daily manual review of magnetic resonance images for possible artefacts can be time-consuming, so automated methods for computer-assisted quality assessment of magnetic resonance imaging need to be developed. The aim of this study was to develop automatic detection of Gibbs artefacts in magnetic resonance imaging using a deep learning method called transfer learning, and to demonstrate the potential of this approach for the development of an automatic quality control tool for the detection of such artefacts in magnetic resonance imaging. The magnetic resonance image dataset of the scanned phantom for quality assurance was created using a turbo spin-echo pulse sequence in the transverse plane. Images were created to include Gibbs artefacts of varying intensities. The images were annotated by two independent reviewers. The annotated dataset was used to develop a method for Gibbs artefact detection using the transfer learning approach. The VGG-16, VGG-19, and ResNet-152 convolutional neural networks were used as pre-trained networks for transfer learning and compared using 5-fold cross-validation. All accuracies of the classification models were above 97%, while the AUC values were all above 0.99, confirming the high quality of the constructed models. We show that transfer learning can be successfully used to detect Gibbs artefacts on magnetic resonance images. The main advantages of transfer learning are that it can be applied on small training datasets, the procedures to build the models are not so complicated, and they do not require much computational power. This shows the potential of transfer learning for the more general task of detecting artefacts in magnetic resonance images of patients, which consequently can improve and speed up the process of quality assessment in medical imaging practice.

A variational approach to Gibbs artifacts removal in MRI

Gibbs ringing is a feature of MR images caused by the finite sampling of the acquisition space (k-space). It manifests itself with ringing patterns around sharp edges which become increasingly significant for low-resolution acquisitions. In this paper, we model the Gibbs artefact removal as a constrained variational problem where the data discrepancy, represented in denoising and convolutive form, is balanced to sparsity-promoting regularization functions such as Total Variation, Total Generalized Variation and L_1 norm of the Wavelet transform. The efficacy of such models is evaluated by running a set of numerical experiments both on synthetic data and real acquisitions of brain images. The Total Generalized Variation penalty coupled with the convolutive data discrepancy term yields, in general, the best results both on synthetic and real data.

Automated filter selection for suppression of Gibbs ringing artefacts in MRI

Gibbs ringing creates artefacts in magnetic resonance images that can mislead clinicians. Reconstruction algorithms attempt to suppress Gibbs ringing, or an additional ringing suppression algorithm may be applied post reconstruction. Novel reconstruction algorithms are often compared with filtered Fourier reconstruction, but the choices of filters and filter parameters can be arbitrary and sub-optimal. Evaluation of different reconstruction and post-processing algorithms is difficult to automate or subjective: many metrics have been used in the literature. In this paper, we evaluate twelve of those metrics and demonstrate that none of them are fit for purpose. We propose a novel metric and demonstrate its efficacy in 1D and 2D simulations. We use our new metric to optimise and compare 17 smoothing filters for suppression of Gibbs artefacts. We examine the transfer functions of the optimised filters, with counter-intuitive results regarding the highest-performing filters. Our results will simplify and improve the comparison of novel MRI reconstruction and post-processing algorithms, and lead to the automation of ringing suppression in MRI. They also apply more generally to other applications in which data is captured in the Fourier domain.

Optimization of 99m Tc whole-body SPECT/CT image quality: A phantom study.

PurposeInvestigate the impact of acquisition time and reconstruction parameters on single‐photon emission computed tomography/computed tomography (SPECT/CT) image quality with the ultimate aim of finding the shortest possible acquisition time for clinical whole‐body SPECT/CT (WB‐SPECT/CT) while maintaining image qualityMethodsThe National Electrical Manufacturers Association (NEMA) image quality measurements were performed on a SPECT/CT imaging system using a NEMA International Electrotechnical Commission (IEC) phantom with spherical inserts of varying diameter (10–37 mm), filled with 99mTc in activity sphere‐to‐background concentration ratio of 8.5:1. A gated acquisition was acquired and binned data were summed to simulate acquisitions of 15, 8, and 3 s per projection angle. Images were reconstructed on a Hermes (HERMES Medical Solutions AB, Stockholm, Sweden) workstation using eight subsets and between 4 and 24 iterations of the three‐dimensional (3D) ordered subset expectation maximization (OSEM) algorithm. Reconstructed images were post‐smoothed with 3D Gaussian filter ranging from 0 to 12 mm full‐width at half maximum (FWHM). Contrast recovery, background variability, and contrast‐to‐noise ratio were evaluatedResultsAs expected, the spheres were more clearly defined as acquisition time and count statistics improved. The optimal iteration number and Gaussian filter were determined from the contrast recovery convergence and level of noise. Convergence of contrast recovery was observed at eight iterations while 12 iterations yielded stabilized values at all acquisition times. In addition, it was observed that applying 3D Gaussian filter of 8–12 mm FWHM suppressed the noise and mitigated Gibbs artifacts. Background variability was larger for small spheres than larger spheres and the noise decreased when acquisition time became longer. A contrast‐to‐noise ratio >5 was reached for the two smallest spheres of 10 and 13 mm at acquisition times of 8 sConclusionOptimized reconstruction parameters preserved image quality with reduce acquisition time in present study. This study suggests an optimal protocol for clinical 99mTc SPECT/CT can be reached at 8 s per projection angle, with data reconstructed using 12 iterations and eight subset of the 3D OSEM algorithm and 8 mm Gaussian post‐filter.

Gibbs or Truncation Artifact on MRI Mimicking Degenerative Cervical Myelopathy.

Gibbs or Truncation Artifact on MRI Mimicking Degenerative Cervical Myelopathy.

Comparative study of different Gd-EOB-DTPA flows as a solution of the Gibbs Arterial Phase artifact in liver MRI

The purpose of this study was to compare three different injection rates to improve the quality of the arterial phaseimages in the MRI imaging of the liver with Gadoxetico Acid (Gd-EOB-DTPA) and, in particular, how these affectand/or mitigate Gibbs artifact.

Comparative study of different Gd-EOB-DTPA flows as a solution of the Gibbs Arterial Phase artifact in liver MRI

The purpose of this study was to compare three different injection rates to improve the quality of the arterial phase images in the MRI imaging of the liver with Gadoxetico Acid (Gd-EOB-DTPA) and, in particular, how these affect and/or mitigate Gibbs artifact.

Seismic denoising via truncated nuclear norm minimization

Because there are many similar geologic structures underground, seismic profiles have an abundance of self-repeating patterns. Thus, we can divide a seismic profile into groups of blocks with a similar seismic structure. The matrix formed by stacking together similar blocks in each group should be of low rank. Hence, we can transfer the seismic denoising problem to a series of low-rank matrix approximation (LRMA) problems. The LRMA-based model commonly adopts the nuclear norm as a convex substitute of the rank of a matrix. However, the nuclear norm minimization (NNM) shrinks the different rank components equally and may cause some biases in practice. The recently introduced truncated nuclear norm (TNN) has been proven to more accurately approximate the rank of a matrix, which is given by the sum of the set of the smallest singular values. Based on this, we have adopted a novel denoising method using truncated nuclear norm minimization (TNNM). The objective function of this method consists of two terms, the F-norm data fidelity and a TNN regularization. We develop an efficient two-step iterative algorithm to solve this objective function. Then, we apply the proposed TNNM algorithm to groups of blocks with similar seismic structure, and we aggregate all resulting denoised blocks to get the denoised seismic data. We update the denoised results during each iteration to gradually attenuate the heavy noise. Numerical experiments demonstrate that, compared with f-x deconvolution and the curvelet- and NNM-based methods, TNNM not only attenuates noise more effectively even when the signal-to-noise ratio is as low as −10 dB and seismic data have complex structures, but it also accurately preserves the seismic structures without inducing Gibbs artifacts.

Theranostic SPECT reconstruction for improved resolution: application to radionuclide therapy dosimetry

BackgroundSPECT-derived dose estimates in tissues of diameter less than 3× system resolution are subject to significant losses due to the limited spatial resolution of the gamma camera. Incorporating resolution modelling (RM) into the SPECT reconstruction has been proposed as a possible solution; however, the images produced are prone to noise amplification and Gibbs artefacts. We propose a novel approach to SPECT reconstruction in a theranostic setting, which we term SPECTRE (single photon emission computed theranostic reconstruction); using a diagnostic PET image, with its superior resolution, to guide the SPECT reconstruction of the therapeutic equivalent. This report demonstrates a proof in principle of this approach.MethodsWe have employed the hybrid kernelised expectation maximisation (HKEM) algorithm implemented in STIR, with the aim of producing SPECT images with PET-equivalent resolution. We demonstrate its application in both a dual 68Ga/177Lu IEC phantom study and a clinical example using 64Cu/67Cu.ResultsSPECTRE is shown to produce images comparable in accuracy and recovery to PET with minimal introduction of artefacts and amplification of noise.ConclusionThe SPECTRE approach to image reconstruction shows improved quantitative accuracy with a reduction in noise amplification. SPECTRE shows great promise as a method of improving SPECT radioactivity concentrations, directly leading to more accurate dosimetry estimates in small structures and target lesions. Further investigation and optimisation of the algorithm parameters is needed before this reconstruction method can be utilised in a clinical setting.

Augmented Noise Learning Framework for Enhancing Medical Image Denoising

Deep learning attempts medical image denoising either by directly learning the noise present or via first learning the image content. We observe that residual learning (RL) often suffers from signal leakage while dictionary learning (DL) is prone to Gibbs (ringing) artifacts. In this paper, we propose an unsupervised noise learning framework that enhances denoising by augmenting the limitation of RL with the strength of DL and vice versa. To this end, we propose a ten-layer deep residue network (DRN) augmented with patch-based dictionaries. The input images are presented to patch-based DL to indirectly learn the noise via sparse representation while given to the DRN to directly learn the noise. An optimum noise characterization is captured by iterating DL/DRN network against proposed loss. The denoised images are obtained by subtracting the learned noise from available data. We show that augmented DRN effectively handles high-frequency regions to avoid Gibbs artifacts due to DL while augmented DL helps to reduce the overfitting due to RL. Comparative experiments with many state-of-the-arts on MRI and CT datasets (2D/3D) including low-dose CT (LDCT) are conducted on a GPU-based supercomputer. The proposed network is trained by adding different levels of Rician noise for MRI and Poisson noise for CT images considering different nature and statistical distribution of datasets. The ablation studies are carried out that demonstrate enhanced denoising performance with minimal signal leakage and least artifacts by proposed augmented approach.

Projection onto the convex sets model based on non‐downsampling contourlet transform and high‐frequency iteration

Obtaining high-resolution images is one of the critical issues in many areas. The super-resolution (SR) algorithm is a method of obtaining a high-resolution image from low-resolution images. The projection onto the convex sets (POCS) iteration method is an essential spatial multi-image SR algorithm. In addition to generating Gibbs artefacts, the traditional POCS algorithm has some restrictions on the image source, such as the signal-to-noise ratio and the resolution of the reference frame. In order to remove the limitation of image source, eliminate artefact, and get more edge information from iteration. This Letter proposes a new POCS iterative model. This Letter uses the high-frequency (HF) image obtained after a non-downsampling contourlet transform as the iteration object. It uses the unique attributes of the HF image to establish a new constrained convex set to construct the iterative formula. In order to make the iteration effect more apparent, this Letter has Laplace enhancement to the reference frame selected by principal component analysis. The experimental results show that the results obtained by the proposed algorithm are satisfactory in both subjective and objective evaluations.

Training a neural network for Gibbs and noise removal in diffusion MRI.

To develop and evaluate a neural network-based method for Gibbs artifact and noise removal. A convolutional neural network (CNN) was designed for artifact removal in diffusion-weighted imaging data. Two implementations were considered: one for magnitude images and one for complex images. Both models were based on the same encoder-decoder structure and were trained by simulating MRI acquisitions on synthetic non-MRI images. Both machine learning methods were able to mitigate artifacts in diffusion-weighted images and diffusion parameter maps. The CNN for complex images was also able to reduce artifacts in partial Fourier acquisitions. The proposed CNNs extend the ability of artifact correction in diffusion MRI. The machine learning method described here can be applied on each imaging slice independently, allowing it to be used flexibly in clinical applications.

Performance evaluation of a new time of flight PET/CT scanner: Results of a multicenter study

PurposeThe aim of this multicenter study was to evaluate the performance of the upgraded version of the Ingenuity TF PET/CT scanner, according to the NEMA NU-2 2012 standards. MethodsSpatial resolution, sensitivity, count rate response, scatter fraction, image quality and accuracy were evaluated on three Ingenuity TF scanners installed in Italian hospitals. Furthermore, energy and timing resolution were measured. A detailed image quality phantom analysis was performed to evaluate the effect of different clinical reconstruction parameters, including the application of PSF correction. ResultsResults show an average spatial resolution of 4.7 mm and an average absolute system sensitivity of 7.9 cps/kBq. The average maximum NECR was 119.83 kcps at 20.67 kBq/ml, while the maximum true event rate was 322.62 kcps at the concentration of 24.51 kBq/ml. The average maximum bias below NECR peak was 12.58%. All the results of NEMA tests were in agreement with the values declared by the manufacturer. The estimated average energy and timing resolution were 10.83% and 536.2 ps, respectively.Image quality phantom analysis obtained with different reconstruction settings showed that PSF correction was the parameter that affected mainly on contrast recovery coefficient, while the iteration number and amplitude of Gaussian filter had no significant effect. Of relevance, the application of PSF correction never led to recovery coefficient values higher than 100% and to Gibbs or edge artifacts. ConclusionsThe new Ingenuity TF model shows physical performance similar to other scanners of the latest generation for all standard NEMA NU2-2012 measurements.

Elimination of Gibbs artifact based on local subpixel shift and interlaced local variation

To extend the application of Gibbs artifact reduction method that exploits local subvoxel- shifts (LSS) to zero- padded k-space magnetic resonance imaging (MRI) data. We investigated two approaches to extending the application of LSS-based method to under-sampled data. The first approach, namely LSS+ interpolation, utilized the original LSS-based method to minimize the local variation on nonzero-padding reconstructed images, followed by image interpolation to obtain the final images. The second approach, interlaced local variation, used zero-padded Fourier transformation followed by elimination of Gibbs artifacts by minimizing a novel interlaced local variations (iLV) term. We compared the two methods with the original LSS and Hamming window filter algorithms, and verified their feasibility and robustness in phantom and in vivo data. The two methods proposed showed better performance than the original LSS and Hamming window filters and effectively eliminated Gibbs artifacts while preserving the image details. Compared to LSS + interpolation method, iLV method better preserved the details of the images. The iLV and LSS+interpolation methods proposed herein both extend the application of the original LSS method and can eliminate Gibbs artifacts in zero-filled k-space data reconstruction images, and iLV method shows a more prominent advantage in retaining the image details.