High-field mr diffusion-weighted image denoising using a joint denoising convolutional neural network

Self-supervised structural similarity-based convolutional neural network for cardiac diffusion tensor image denoising

BACKGROUND

Diffusion tensor imaging (DTI) is a promising technique for non-invasively investigating the myocardial fiber structures of human heart. However, low signal-to-noise ratio (SNR) has been a major limit of cardiac DTI to prevent us from detecting myocardium structure accurately. Therefore, it is important to remove the effect of noise on diffusion weighted (DW) images.

PURPOSE

Although the conventional and deep learning-based denoising methods have shown the potential to deal with effectively the noise in DW images, most of them are redundant information dependent or require the noise-free images as golden standard. In addition, the existed DW image denoising methods often suffer from problems of over-smoothing. To address these issues, we propose a self-supervised learning model, structural similarity based convolutional neural network with edge-weighted loss (SSECNN), to remove the noise effectively in cardiac DTI.

METHODS

Considering that the DW images acquired along different diffusion directions have structural similarity, and the noise in these DW images is independent and identically distributed, the structural similarity-based matching algorithm is proposed to search for the most similar DW images. Such similar noisy DW image pairs are then used as the input and target of the denoising network SSECNN, which consists of several convolutional and residual blocks. Through the self-supervised training with these image pairs, the network can restore the clean DW images and retain the correlations between the denoised DW images along different directions. To avoid the over-smoothing problem, we design a novel edge-weighted loss which enables the network to adaptively adjust the loss weights with iterations and therefore to improve the detail preserve ability of the model. To verify the superiority of the proposed method, comparisons with state-of-the-art (SOTA) denoising methods are performed on both synthetic and real acquired DTI datasets.

RESULTS

Experimental results show that SSECNN can effectively reduce the noise in the DW images while preserving detailed texture and edge information and therefore achieve better performance in DTI reconstruction. For synthetic dataset, compared to the SOTA method, the root mean square error (RMSE), peak signal-to-noise ratio (PSNR), and structure similarity index measure (SSIM) between the denoised DW images obtained with SSECNN and noise-free DW images are improved by 6.94%, 1.98%, and 0.76% respectively when the noise level is 10%. As for the acquired cardiac DTI dataset, the SSECNN method could significantly improve SNR and contrast to noise ratio (CNR) of cardiac DW images and achieve more regular helix angle (HA) and transverse angle (TA) maps. The ablation experimental results validate that using the structure similarity-based method to search the similar DW image pairs yield the smallest loss, and with the help of the edge-weighted loss, the denoised DW images and diffusion metric maps can preserve more details.

CONCLUSIONS

The proposed SSECNN method can fully explore the similarity between the DW images along different diffusion directions. Using such similarity and an edge-weighted loss enable us to denoise cardiac DTI effectively in a self-supervised manner. Our method can overcome the redundancy information dependence and over-smoothing problem of the SOTA methods.

Magnetic resonance imaging and ultrasound elastography in the context of preclinical pharmacological research: significance for the 3R principles

The 3Rs principles-reduction, refinement, replacement-are at the core of preclinical research within drug discovery, which still relies to a great extent on the availability of models of disease in animals. Minimizing their distress, reducing their number as well as searching for means to replace them in experimental studies are constant objectives in this area. Due to its non-invasive character in vivo imaging supports these efforts by enabling repeated longitudinal assessments in each animal which serves as its own control, thereby enabling to reduce considerably the animal utilization in the experiments. The repetitive monitoring of pathology progression and the effects of therapy becomes feasible by assessment of quantitative biomarkers. Moreover, imaging has translational prospects by facilitating the comparison of studies performed in small rodents and humans. Also, learnings from the clinic may be potentially back-translated to preclinical settings and therefore contribute to refining animal investigations. By concentrating on activities around the application of magnetic resonance imaging (MRI) and ultrasound elastography to small rodent models of disease, we aim to illustrate how in vivo imaging contributes primarily to reduction and refinement in the context of pharmacological research.

Accelerated Diffusion-Weighted MRI of Rectal Cancer Using a Residual Convolutional Network

This work presents a deep-learning-based denoising technique to accelerate the acquisition of high b-value diffusion-weighted MRI for rectal cancer. A denoising convolutional neural network (DCNN) with a combined L1–L2 loss function was developed to denoise high b-value diffusion-weighted MRI data acquired with fewer repetitions (NEX: number of excitations) using the low b-value image as an anatomical guide. DCNN was trained using 85 datasets acquired on patients with rectal cancer and tested on 20 different datasets with NEX = 1, 2, and 4, corresponding to acceleration factors of 16, 8, and 4, respectively. Image quality was assessed qualitatively by expert body radiologists. Reader 1 scored similar overall image quality between denoised images with NEX = 1 and NEX = 2, which were slightly lower than the reference. Reader 2 scored similar quality between NEX = 1 and the reference, while better quality for NEX = 2. Denoised images with fourfold acceleration (NEX = 4) received even higher scores than the reference, which is due in part to the effect of gas-related motion in the rectum, which affects longer acquisitions. The proposed deep learning denoising technique can enable eightfold acceleration with similar image quality (average image quality = 2.8 ± 0.5) and fourfold acceleration with higher image quality (3.0 ± 0.6) than the clinical standard (2.5 ± 0.8) for improved diagnosis of rectal cancer.

Convolutional Neural Network for Denoising Left Ventricle Magnetic Resonance Images

Medical image processing is critical in disease detection and prediction. For example, they locate lesions and measure an organ's morphological structures. Currently, cardiac magnetic resonance imaging (CMRI) plays an essential role in cardiac motion tracking and analyzing regional and global heart functions with high accuracy and reproducibility. Cardiac MRI datasets are images taken during the heart's cardiac cycles. These datasets require expert labeling to accurately recognize features and train neural networks to predict cardiac disease. Any erroneous prediction caused by image impairment will impact patients' diagnostic decisions. As a result, image preprocessing is used, including enhancement tools such as filtering and denoising. This paper introduces a denoising algorithm that uses a convolution neural network (CNN) to delineate left ventricle (LV) contours (endocardium and epicardium borders) from MRI images. With only a small amount of training data from the EMIDEC database, this network performs well for MRI image denoising.

Accelerate gas diffusion-weighted MRI for lung morphometry with deep learning

OBJECTIVES

Multiple b-value gas diffusion-weighted MRI (DW-MRI) enables non-invasive and quantitative assessment of lung morphometry, but its long acquisition time is not well-tolerated by patients. We aimed to accelerate multiple b-value gas DW-MRI for lung morphometry using deep learning.

METHODS

A deep cascade of residual dense network (DC-RDN) was developed to reconstruct high-quality DW images from highly undersampled k-space data. Hyperpolarized ¹²⁹Xe lung ventilation images were acquired from 101 participants and were retrospectively collected to generate synthetic DW-MRI data to train the DC-RDN. Afterwards, the performance of the DC-RDN was evaluated on retrospectively and prospectively undersampled multiple b-value ¹²⁹Xe MRI datasets.

RESULTS

Each slice with size of 64 × 64 × 5 could be reconstructed within 7.2 ms. For the retrospective test data, the DC-RDN showed significant improvement on all quantitative metrics compared with the conventional reconstruction methods (p < 0.05). The apparent diffusion coefficient (ADC) and morphometry parameters were not significantly different between the fully sampled and DC-RDN reconstructed images (p > 0.05). For the prospectively accelerated acquisition, the required breath-holding time was reduced from 17.8 to 4.7 s with an acceleration factor of 4. Meanwhile, the prospectively reconstructed results showed good agreement with the fully sampled images, with a mean difference of -0.72% and -0.74% regarding global mean ADC and mean linear intercept (L_m) values.

CONCLUSIONS

DC-RDN is effective in accelerating multiple b-value gas DW-MRI while maintaining accurate estimation of lung microstructural morphometry, facilitating the clinical potential of studying lung diseases with hyperpolarized DW-MRI.

KEY POINTS

• The deep cascade of residual dense network allowed fast and high-quality reconstruction of multiple b-value gas diffusion-weighted MRI at an acceleration factor of 4. • The apparent diffusion coefficient and morphometry parameters were not significantly different between the fully sampled images and the reconstructed results (p > 0.05). • The required breath-holding time was reduced from 17.8 to 4.7 s and each slice with size of 64 × 64 × 5 could be reconstructed within 7.2 ms.

Neural networks for parameter estimation in microstructural MRI: Application to a diffusion-relaxation model of white matter

Specific features of white matter microstructure can be investigated by using biophysical models to interpret relaxation-diffusion MRI brain data. Although more intricate models have the potential to reveal more details of the tissue, they also incur time-consuming parameter estimation that may converge to inaccurate solutions due to a prevalence of local minima in a degenerate fitting landscape. Machine-learning fitting algorithms have been proposed to accelerate the parameter estimation and increase the robustness of the attained estimates. So far, learning-based fitting approaches have been restricted to microstructural models with a reduced number of independent model parameters where dense sets of training data are easy to generate. Moreover, the degree to which machine learning can alleviate the degeneracy problem is poorly understood. For conventional least-squares solvers, it has been shown that degeneracy can be avoided by acquisition with optimized relaxation-diffusion-correlation protocols that include tensor-valued diffusion encoding. Whether machine-learning techniques can offset these acquisition requirements remains to be tested. In this work, we employ artificial neural networks to vastly accelerate the parameter estimation for a recently introduced relaxation-diffusion model of white matter microstructure. We also develop strategies for assessing the accuracy and sensitivity of function fitting networks and use those strategies to explore the impact of the acquisition protocol. The developed learning-based fitting pipelines were tested on relaxation-diffusion data acquired with optimal and sub-optimal acquisition protocols. Networks trained with an optimized protocol were observed to provide accurate parameter estimates within short computational times. Comparing neural networks and least-squares solvers, we found the performance of the former to be less affected by sub-optimal protocols; however, model fitting networks were still susceptible to degeneracy issues and their use could not fully replace a careful design of the acquisition protocol.

Quantitative diffusion MRI of the abdomen and pelvis

Diffusion MRI has enormous potential and utility in the evaluation of various abdominal and pelvic disease processes including cancer and noncancer imaging of the liver, prostate, and other organs. Quantitative diffusion MRI is based on acquisitions with multiple diffusion encodings followed by quantitative mapping of diffusion parameters that are sensitive to tissue microstructure. Compared to qualitative diffusion-weighted MRI, quantitative diffusion MRI can improve standardization of tissue characterization as needed for disease detection, staging, and treatment monitoring. However, similar to many other quantitative MRI methods, diffusion MRI faces multiple challenges including acquisition artifacts, signal modeling limitations, and biological variability. In abdominal and pelvic diffusion MRI, technical acquisition challenges include physiologic motion (respiratory, peristaltic, and pulsatile), image distortions, and low signal-to-noise ratio. If unaddressed, these challenges lead to poor technical performance (bias and precision) and clinical outcomes of quantitative diffusion MRI. Emerging and novel technical developments seek to address these challenges and may enable reliable quantitative diffusion MRI of the abdomen and pelvis. Through systematic validation in phantoms, volunteers, and patients, including multicenter studies to assess reproducibility, these emerging techniques may finally demonstrate the potential of quantitative diffusion MRI for abdominal and pelvic imaging applications.

[A diffusion-weighted image denoising algorithm using HOSVD combined with Rician noise corrected model]

OBJECTIVE

To propose a novel diffusion-weighted (DW) image denoising algorithm based on HOSVD to improve the signal-to-noise ratio (SNR) of DW images and the accuracy of subsequent quantization parameters.

METHODS

This HOSVDbased denoising method incorporated the sparse constraint and noise-correction model. The signal expectations with Rician noise were integrated into the traditional HOSVD denoising framework for direct denoising of the DW images with Rician noise. HOSVD denoising was performed directly on each local DW image block to avoid the stripe artifacts. We compared the proposed method with 4 image denoising algorithms (LR + Edge, GL-HOSVD, BM3D and NLM) to verify the effect of the proposed method.

RESULTS

The experimental results showed that the proposed method effectively reduced the noise of DW images while preserving the image details and edge structure information. The proposed algorithm was significantly better than LR +Edge, BM3D and NLM in terms of quantitative metrics of PSNR, SSIM and FA-RMSE and in visual evaluation of denoising images and FA images. GL-HOSVD obtained good denoising results but introduced stripe artifacts at a high noise level during the denoising process. In contrast, the proposed method achieved good denoising results without causing stripe artifacts.

CONCLUSION

This HOSVD-based denoising method allows direct processing of DW images with Rician noise without introducing artifacts and can provide accurate quantitative parameters for diagnostic purposes.

Continuous diffusion spectrum computation for diffusion-weighted magnetic resonance imaging of the kidney tubule system

Background

The use of rigid multi-exponential models (with a priori predefined numbers of components) is common practice for diffusion-weighted MRI (DWI) analysis of the kidney. This approach may not accurately reflect renal microstructure, as the data are forced to conform to the a priori assumptions of simplified models. This work examines the feasibility of less constrained, data-driven non-negative least squares (NNLS) continuum modelling for DWI of the kidney tubule system in simulations that include emulations of pathophysiological conditions.

Methods

Non-linear least squares (LS) fitting was used as reference for the simulations. For performance assessment, a threshold of 5% or 10% for the mean absolute percentage error (MAPE) of NNLS and LS results was used. As ground truth, a tri-exponential model using defined volume fractions and diffusion coefficients for each renal compartment (tubule system: D_tubules , f_tubules ; renal tissue: D_tissue , f_tissue ; renal blood: D_blood , f_blood ;) was applied. The impact of: (I) signal-to-noise ratio (SNR) =40-1,000, (II) number of b-values (n=10-50), (III) diffusion weighting (b-range_small =0-800 up to b-range_large =0-2,180 s/mm²), and (IV) fixation of the diffusion coefficients D_tissue and D_blood was examined. NNLS was evaluated for baseline and pathophysiological conditions, namely increased tubular volume fraction (ITV) and renal fibrosis (10%: grade I, mild) and 30% (grade II, moderate).

Results

NNLS showed the same high degree of reliability as the non-linear LS. MAPE of the tubular volume fraction (f_tubules ) decreased with increasing SNR. Increasing the number of b-values was beneficial for f_tubules precision. Using the b-range_large led to a decrease in MAPE _ftubules compared to b-range_small. The use of a medium b-value range of b=0-1,380 s/mm² improved f_tubules precision, and further b_max increases beyond this range yielded diminishing improvements. Fixing D_blood and D_tissue significantly reduced MAPE _ftubules and provided near perfect distinction between baseline and ITV conditions. Without constraining the number of renal compartments in advance, NNLS was able to detect the (fourth) fibrotic compartment, to differentiate it from the other three diffusion components, and to distinguish between 10% vs. 30% fibrosis.

Conclusions

This work demonstrates the feasibility of NNLS modelling for DWI of the kidney tubule system and shows its potential for examining diffusion compartments associated with renal pathophysiology including ITV fraction and different degrees of fibrosis.

Applicability of deep learning-based reconstruction trained by brain and knee 3T MRI to lumbar 1.5T MRI

Background

Several deep learning-based methods have been proposed for addressing the long scanning time of magnetic resonance imaging. Most are trained using brain 3T magnetic resonance images, but is unclear whether performance is affected when applying these methods to different anatomical sites and at different field strengths.

Purpose

To validate the denoising performance of deep learning-based reconstruction method trained by brain and knee 3T magnetic resonance images when applied to lumbar 1.5T magnetic resonance images.

Material and Methods

Using a 1.5T scanner, we obtained lumber T2-weighted sequences in 10 volunteers using three different scanning times: 228 s (standard), 119 s (double-fast), and 68 s (triple-fast). We compared the images obtained by the standard sequence with those obtained by the deep learning-based reconstruction-applied faster sequences.

Results

Signal-to-noise ratio values were significantly higher for deep learning-based reconstruction-double-fast than for standard and did not differ significantly between deep learning-based reconstruction-triple-fast and standard. Contrast-to-noise ratio values also did not differ significantly between deep learning-based reconstruction-triple-fast and standard. Qualitative scores for perceived signal-to-noise ratio and overall image quality were significantly higher for deep learning-based reconstruction-double fast and deep learning-based reconstruction-triple-fast than for standard. Average scores for sharpness, contrast, and structure visibility were equal to or higher for deep learning-based reconstruction-double-fast and deep learning-based reconstruction-triple-fast than for standard, but the differences were not statistically significant. The average scores for artifact were lower for deep learning-based reconstruction-double-fast and deep learning-based reconstruction-triple-fast than for standard, but the differences were not statistically significant.

Conclusion

The deep learning-based reconstruction method trained by 3T brain and knee images may reduce the scanning time of 1.5T lumbar magnetic resonance images by one-third without sacrificing image quality.

Deep Learning-Based Denoising for High b-Value at 2000 s/mm2 Diffusion-Weighted Imaging

Diffusion-weighted imaging (DWI) allows white matter quantification of the white matter tracts of the brain. However, at a high b-value (≥ 2000 s/mm2), DWI acquisition suffers from noise due to longer acquisition times obscuring white matter interpretation. DWI denoising techniques can be used to acquire high b-value DWI without increasing the number of signal averages. We used a residual learning-based convolutional neural network (DnCNN) to reduce noise in high b-value DWI based on the literature review. We applied the proposed denoising method on high b-value, retrospectively collected DWI data with multiple noise levels. Experimental results show an improved image quality after denoising in retrospective DWI (average PSNR before and after denoising: 27.63 ± 1.06 dB and 51.76 ± 1.95 dB, respectively). The prospective DWI included one and two signal averages for denoising. DWI with four signal averages was used as the reference. Representative images show high b-value prospective DW images denoised using the DnCNN. We demonstrated DnCNN for cases of multiple noise levels and signal averages. For the prospective study, the PSNR values for 1-NEX before and after denoising were 27.39 ± 3.75 dB and 27.68 ± 3.75 dB. For 2-NEX, the PSNR values before and after denoising were 27.51 ± 4.18 dB and 27.75 ± 4.05 dB.

Accelerated Acquisition of High-resolution Diffusion-weighted Imaging of the Brain with a Multi-shot Echo-planar Sequence: Deep-learning-based Denoising

To accelerate high-resolution diffusion-weighted imaging with a multi-shot echo-planar sequence, we propose an approach based on reduced averaging and deep learning. Denoising convolutional neural networks can reduce amplified noise without requiring extensive averaging, enabling shorter scan times and high image quality. The preliminary experimental results demonstrate the superior performance of the proposed denoising method over state-of-the-art methods such as the widely used block-matching and 3D filtering.

Transferring principles of solid-state and Laplace NMR to the field of in vivo brain MRI

Magnetic resonance imaging (MRI) is the primary method for noninvasive investigations of the human brain in health, disease, and development but yields data that are difficult to interpret whenever the millimeter-scale voxels contain multiple microscopic tissue environments with different chemical and structural properties. We propose a novel MRI framework to quantify the microscopic heterogeneity of the living human brain as spatially resolved five-dimensional relaxation-diffusion distributions by augmenting a conventional diffusion-weighted imaging sequence with signal encoding principles from multidimensional solid-state nuclear magnetic resonance (NMR) spectroscopy, relaxation-diffusion correlation methods from Laplace NMR of porous media, and Monte Carlo data inversion. The high dimensionality of the distribution space allows resolution of multiple microscopic environments within each heterogeneous voxel as well as their individual characterization with novel statistical measures that combine the chemical sensitivity of the relaxation rates with the link between microstructure and the anisotropic diffusivity of tissue water. The proposed framework is demonstrated on a healthy volunteer using both exhaustive and clinically viable acquisition protocols.