Denoising of 3D magnetic resonance images with multi-channel residual learning of convolutional neural network

Fast reconstruction of SMS bSSFP myocardial perfusion images using noise map estimation network (NoiseMapNet): a head-to-head comparison with parallel imaging and iterative reconstruction

Background

Simultaneous multi-slice (SMS) bSSFP imaging enables stress myocardial perfusion imaging with high spatial resolution and increased spatial coverage. Standard parallel imaging techniques (e.g., TGRAPPA) can be used for image reconstruction but result in high noise level. Alternatively, iterative reconstruction techniques based on temporal regularization (ITER) improve image quality but are associated with reduced temporal signal fidelity and long computation time limiting their online use. The aim is to develop an image reconstruction technique for SMS-bSSFP myocardial perfusion imaging combining parallel imaging and image-based denoising using a novel noise map estimation network (NoiseMapNet), which preserves both sharpness and temporal signal profiles and that has low computational cost.

Methods

The proposed reconstruction of SMS images consists of a standard temporal parallel imaging reconstruction (TGRAPPA) with motion correction (MOCO) followed by image denoising using NoiseMapNet. NoiseMapNet is a deep learning network based on a 2D Unet architecture and aims to predict a noise map from an input noisy image, which is then subtracted from the noisy image to generate the denoised image. This approach was evaluated in 17 patients who underwent stress perfusion imaging using a SMS-bSSFP sequence. Images were reconstructed with (a) TGRAPPA with MOCO (thereafter referred to as TGRAPPA), (b) iterative reconstruction with integrated motion compensation (ITER), and (c) proposed NoiseMapNet-based reconstruction. Normalized mean squared error (NMSE) with respect to TGRAPPA, myocardial sharpness, image quality, perceived SNR (pSNR), and number of diagnostic segments were evaluated.

Results

NMSE of NoiseMapNet was lower than using ITER for both myocardium (0.045 ± 0.021 vs. 0.172 ± 0.041, p < 0.001) and left ventricular blood pool (0.025 ± 0.014 vs. 0.069 ± 0.020, p < 0.001). There were no significant differences between all methods for myocardial sharpness (p = 0.77) and number of diagnostic segments (p = 0.36). ITER led to higher image quality than NoiseMapNet/TGRAPPA (2.7 ± 0.4 vs. 1.8 ± 0.4/1.3 ± 0.6, p < 0.001) and higher pSNR than NoiseMapNet/TGRAPPA (3.0 ± 0.0 vs. 2.0 ± 0.0/1.3 ± 0.6, p < 0.001). Importantly, NoiseMapNet yielded higher pSNR (p < 0.001) and image quality (p < 0.008) than TGRAPPA. Computation time of NoiseMapNet was only 20s for one entire dataset.

Conclusion

NoiseMapNet-based reconstruction enables fast SMS image reconstruction for stress myocardial perfusion imaging while preserving sharpness and temporal signal profiles.

Rapid 2D 23Na MRI of the calf using a denoising convolutional neural network

PURPOSE

23Na MRI can be used to quantify in-vivo tissue sodium concentration (TSC), but the inherently low 23Na signal leads to long scan times and/or noisy or low-resolution images. Reconstruction algorithms such as compressed sensing (CS) have been proposed to mitigate low signal-to-noise ratio (SNR); although, these can result in unnatural images, suboptimal denoising and long processing times. Recently, machine learning has been increasingly used to denoise 1H MRI acquisitions; however, this approach typically requires large volumes of high-quality training data, which is not readily available for 23Na MRI. Here, we propose using 1H data to train a denoising convolutional neural network (CNN), which we subsequently demonstrate on prospective 23Na images of the calf.

METHODS

1893 1H fat-saturated transverse slices of the knee from the open-source fastMRI dataset were used to train denoising CNNs for different levels of noise. Synthetic low SNR images were generated by adding gaussian noise to the high-quality 1H k-space data before reconstruction to create paired training data. For prospective testing, 23Na images of the calf were acquired in 10 healthy volunteers with a total of 150 averages over ten minutes, which were used as a reference throughout the study. From this data, images with fewer averages were retrospectively reconstructed using a non-uniform fast Fourier transform (NUFFT) as well as CS, with the NUFFT images subsequently denoised using the trained CNN.

RESULTS

CNNs were successfully applied to 23Na images reconstructed with 50, 40 and 30 averages. Muscle and skin apparent TSC quantification from CNN-denoised images were equivalent to those from CS images, with <0.9 mM bias compared to reference values. Estimated SNR was significantly higher in CNN-denoised images compared to NUFFT, CS and reference images. Quantitative edge sharpness was equivalent for all images. For subjective image quality ranking, CNN-denoised images ranked equally best with reference images and significantly better than NUFFT and CS images.

CONCLUSION

Denoising CNNs trained on 1H data can be successfully applied to 23Na images of the calf; thus, allowing scan time to be reduced from ten minutes to two minutes with little impact on image quality or apparent TSC quantification accuracy.

Effect of deep learning-based reconstruction on high-resolution three-dimensional T2-weighted fast asymmetric spin-echo imaging in the preoperative evaluation of cerebellopontine angle tumors

PURPOSE

We aimed to evaluate the effect of deep learning-based reconstruction (DLR) on high-spatial-resolution three-dimensional T2-weighted fast asymmetric spin-echo (HR-3D T2-FASE) imaging in the preoperative evaluation of cerebellopontine angle (CPA) tumors.

METHODS

This study included 13 consecutive patients who underwent preoperative HR-3D T2-FASE imaging using a 3 T MRI scanner. The reconstruction voxel size of HR-3D T2-FASE imaging was 0.23 × 0.23 × 0.5 mm. The contrast-to-noise ratios (CNRs) of the structures were compared between HR-3D T2-FASE images with and without DLR. The observers' preferences based on four categories on the tumor side on HR-3D T2-FASE images were evaluated. The facial nerve in relation to the tumor on HR-3D T2-FASE images was assessed with reference to intraoperative findings.

RESULTS

The mean CNR between the tumor and trigeminal nerve and between the cerebrospinal fluid and trigeminal nerve was significantly higher for DLR images than non-DLR-based images (14.3 ± 8.9 vs. 12.0 ± 7.6, and 66.4 ± 12.0 vs. 53.9 ± 8.5, P < 0.001, respectively). The observer's preference for the depiction and delineation of the tumor, cranial nerves, vessels, and location relation on DLR HR-3D T2FASE images was superior to that on non-DLR HR-3D T2FASE images in 7 (54%), 6 (46%), 6 (46%), and 6 (46%) of 13 cases, respectively. The facial nerves around the tumor on HR-3D T2-FASE images were visualized accurately in five (38%) cases with DLR and in four (31%) without DLR.

CONCLUSION

DLR HR-3D T2-FASE imaging is useful for the preoperative assessment of CPA tumors.

Usefulness of pituitary high-resolution 3D MRI with deep-learning-based reconstruction for perioperative evaluation of pituitary adenomas

PURPOSE

To evaluate the diagnostic value of T1-weighted 3D fast spin-echo sequence (CUBE) with deep learning-based reconstruction (DLR) for depiction of pituitary adenoma and parasellar regions on contrast-enhanced MRI.

METHODS

We evaluated 24 patients with pituitary adenoma or residual tumor using CUBE with and without DLR, 1-mm slice thickness 2D T1WI (1-mm 2D T1WI) with DLR, and 3D spoiled gradient echo sequence (SPGR) as contrast-enhanced MRI. Depiction scores of pituitary adenoma and parasellar regions were assigned by two neuroradiologists, and contrast-to-noise ratio (CNR) was calculated.

RESULTS

CUBE with DLR showed significantly higher scores for depicting pituitary adenoma or residual tumor compared to CUBE without DLR, 1-mm 2D T1WI with DLR, and SPGR (p < 0.01). The depiction score for delineation of the boundary between adenoma and the cavernous sinus was higher for CUBE with DLR than for 1-mm 2D T1WI with DLR (p = 0.01), but the difference was not significant when compared to SPGR (p = 0.20). CUBE with DLR had better interobserver agreement for evaluating adenomas than 1-mm 2D T1WI with DLR (Kappa values, 0.75 vs. 0.41). The CNR of the adenoma to the brain parenchyma increased to a ratio of 3.6 (obtained by dividing 13.7, CNR of CUBE with DLR, by 3.8, that without DLR, p < 0.01). CUBE with DLR had a significantly higher CNR than SPGR, but not 1-mm 2D T1WI with DLR.

CONCLUSION

On the contrast-enhanced MRI, compared to CUBE without DLR, 1-mm 2D T1WI with DLR and SPGR, CUBE with DLR improves the depiction of pituitary adenoma and parasellar regions.

New non-local mean methods for MRI denoising based on global self-similarity between values

Magnetic resonance imaging (MRI) is a non-invasive medical imaging technique that provides high-resolution 3D images and valuable insights into human tissue conditions. Even at present, the refinement of denoising methods for MRI remains a crucial concern for improving the quality of the images. This study aims to improve the prefiltered rotationally invariant non-local principal component analysis (PRI-NL-PCA) algorithm. We relaxed the original restrictions using particle swarm optimization to determine optimal parameters for the PCA part of the original algorithm. In addition, we adjusted the prefiltered rotationally invariant non-local mean (PRI-NLM) part by traversing the signal intensities of voxels instead of their spatial positions to reduce duplicate calculations and expand the search volume to the whole image when estimating voxels' signal intensities. The new method demonstrated superior denoising performance compared to the original approach. Moreover, in most cases, the new algorithm ran faster. Furthermore, our proposed method can also be applied to process Gaussian noise in natural images and has the potential to enhance other NLM-based denoising algorithms.

Flow independent black blood imaging with a large FOV from the neck to the aortic arch: A feasibility study at 3 tesla

PURPOSE

To investigate the feasibility of obtaining black-blood imaging with a large FOV from the neck to the aortic arch at 3 T using a newly modified Relaxation-Enhanced Angiography without Contrast and Triggering for Black-Blood Imaging (REACT-BB).

MATERIALS AND METHODS

REACT-BB provides black-blood images by adjusting the inversion time (TI) in REACT to the null point of blood. The optimal TI for REACT-BB was investigated in 10 healthy volunteers with TI varied from 200 ms to 1400 ms. Contrast ratios were calculated between muscle and three branch arteries of the aortic arch. Additionally, a comparison between REACT-BB and MPRAGE involved evaluating the depiction of high-intensity plaques in 222 patients with stroke or transient ischemic attack. Measurements included plaque-to-muscle signal intensity ratios (PMR), plaque volumes, and carotid artery stenosis rates in 60 patients with high-intensity plaques in carotid arteries.

RESULTS

REACT-BB with TI = 850 ms produced the black-blood image with the best contrast between blood and background tissues. REACT-BB outperformed MPRAGE in depicting high-intensity plaques in the aortic arch (55.4% vs 45.5%) and exhibited superior overall image quality in visual assessment (3.31 ± 0.70 vs 2.89 ± 0.73; p < 0.05). Although the PMR of REACT-BB was significantly lower than MPRAGE (2.227 ± 0.601 vs 2.285 ± 0.662; P < 0.05), a strong positive correlation existed between REACT-BB and MPRAGE (ρ = 0.935; P < 0.05), and all high-intensity plaques that MPRAGE detected were clearly detected by REACT-BB.

CONCLUSION

REACT-BB provides black-blood images with uniformly suppressed fat and blood signals over a large FOV from the neck to the aortic arch with comparable or better high-signal plaque depiction than MPRAGE.

Thin-slice Two-dimensional T2-weighted Imaging with Deep Learning-based Reconstruction: Improved Lesion Detection in the Brain of Patients with Multiple Sclerosis

PURPOSE

Brain MRI with high spatial resolution allows for a more detailed delineation of multiple sclerosis (MS) lesions. The recently developed deep learning-based reconstruction (DLR) technique enables image denoising with sharp edges and reduced artifacts, which improves the image quality of thin-slice 2D MRI. We, therefore, assessed the diagnostic value of 1 mm-slice-thickness 2D T2-weighted imaging (T2WI) with DLR (1 mm T2WI with DLR) compared with conventional MRI for identifying MS lesions.

METHODS

Conventional MRI (5 mm T2WI, 2D and 3D fluid-attenuated inversion recovery) and 1 mm T2WI with DLR (imaging time: 7 minutes) were performed in 42 MS patients. For lesion detection, two neuroradiologists counted the MS lesions in two reading sessions (conventional MRI interpretation with 5 mm T2WI and MRI interpretations with 1 mm T2WI with DLR). The numbers of lesions per region category (cerebral hemisphere, basal ganglia, brain stem, cerebellar hemisphere) were then compared between the two reading sessions.

RESULTS

For the detection of MS lesions by 2 neuroradiologists, the total number of detected MS lesions was significantly higher for MRI interpretation with 1 mm T2WI with DLR than for conventional MRI interpretation with 5 mm T2WI (765 lesions vs. 870 lesions at radiologist A, < 0.05). In particular, of the 33 lesions in the brain stem, radiologist A detected 21 (63.6%) additional lesions by 1 mm T2WI with DLR.

CONCLUSION

Using the DLR technique, whole-brain 1 mm T2WI can be performed in about 7 minutes, which is feasible for routine clinical practice. MRI with 1 mm T2WI with DLR enabled increased MS lesion detection, particularly in the brain stem.

Contrast-enhanced to non-contrast-enhanced image translation to exploit a clinical data warehouse of T1-weighted brain MRI

BACKGROUND

Clinical data warehouses provide access to massive amounts of medical images, but these images are often heterogeneous. They can for instance include images acquired both with or without the injection of a gadolinium-based contrast agent. Harmonizing such data sets is thus fundamental to guarantee unbiased results, for example when performing differential diagnosis. Furthermore, classical neuroimaging software tools for feature extraction are typically applied only to images without gadolinium. The objective of this work is to evaluate how image translation can be useful to exploit a highly heterogeneous data set containing both contrast-enhanced and non-contrast-enhanced images from a clinical data warehouse.

METHODS

We propose and compare different 3D U-Net and conditional GAN models to convert contrast-enhanced T1-weighted (T1ce) into non-contrast-enhanced (T1nce) brain MRI. These models were trained using 230 image pairs and tested on 77 image pairs from the clinical data warehouse of the Greater Paris area.

RESULTS

Validation using standard image similarity measures demonstrated that the similarity between real and synthetic T1nce images was higher than between real T1nce and T1ce images for all the models compared. The best performing models were further validated on a segmentation task. We showed that tissue volumes extracted from synthetic T1nce images were closer to those of real T1nce images than volumes extracted from T1ce images.

CONCLUSION

We showed that deep learning models initially developed with research quality data could synthesize T1nce from T1ce images of clinical quality and that reliable features could be extracted from the synthetic images, thus demonstrating the ability of such methods to help exploit a data set coming from a clinical data warehouse.

Deep learning denoising reconstruction for improved image quality in fetal cardiac cine MRI

Purpose

This study aims to evaluate deep learning (DL) denoising reconstructions for image quality improvement of Doppler ultrasound (DUS)-gated fetal cardiac MRI in congenital heart disease (CHD).

Methods

Twenty-five fetuses with CHD (mean gestational age: 35 ± 1 weeks) underwent fetal cardiac MRI at 3T. Cine imaging was acquired using a balanced steady-state free precession (bSSFP) sequence with Doppler ultrasound gating. Images were reconstructed using both compressed sensing (bSSFP CS) and a pre-trained convolutional neural network trained for DL denoising (bSSFP DL). Images were compared qualitatively based on a 5-point Likert scale (from 1 = non-diagnostic to 5 = excellent) and quantitatively by calculating the apparent signal-to-noise ratio (aSNR) and contrast-to-noise ratio (aCNR). Diagnostic confidence was assessed for the atria, ventricles, foramen ovale, valves, great vessels, aortic arch, and pulmonary veins.

Results

Fetal cardiac cine MRI was successful in 23 fetuses (92%), with two studies excluded due to extensive fetal motion. The image quality of bSSFP DL cine reconstructions was rated superior to standard bSSFP CS cine images in terms of contrast [3 (interquartile range: 2-4) vs. 5 (4-5), P < 0.001] and endocardial edge definition [3 (2-4) vs. 4 (4-5), P < 0.001], while the extent of artifacts was found to be comparable [4 (3-4.75) vs. 4 (3-4), P = 0.40]. bSSFP DL images had higher aSNR and aCNR compared with the bSSFP CS images (aSNR: 13.4 ± 6.9 vs. 8.3 ± 3.6, P < 0.001; aCNR: 26.6 ± 15.8 vs. 14.4 ± 6.8, P < 0.001). Diagnostic confidence of the bSSFP DL images was superior for the evaluation of cardiovascular structures (e.g., atria and ventricles: P = 0.003).

Conclusion

DL image denoising provides superior quality for DUS-gated fetal cardiac cine imaging of CHD compared to standard CS image reconstruction.

Present and Future Innovations in AI and Cardiac MRI

Cardiac MRI is used to diagnose and treat patients with a multitude of cardiovascular diseases. Despite the growth of clinical cardiac MRI, complicated image prescriptions and long acquisition protocols limit the specialty and restrain its impact on the practice of medicine. Artificial intelligence (AI)-the ability to mimic human intelligence in learning and performing tasks-will impact nearly all aspects of MRI. Deep learning (DL) primarily uses an artificial neural network to learn a specific task from example data sets. Self-driving scanners are increasingly available, where AI automatically controls cardiac image prescriptions. These scanners offer faster image collection with higher spatial and temporal resolution, eliminating the need for cardiac triggering or breath holding. In the future, fully automated inline image analysis will most likely provide all contour drawings and initial measurements to the reader. Advanced analysis using radiomic or DL features may provide new insights and information not typically extracted in the current analysis workflow. AI may further help integrate these features with clinical, genetic, wearable-device, and "omics" data to improve patient outcomes. This article presents an overview of AI and its application in cardiac MRI, including in image acquisition, reconstruction, and processing, and opportunities for more personalized cardiovascular care through extraction of novel imaging markers.

Deep learning-based magnetic resonance imaging reconstruction for improving the image quality of reduced-field-of-view diffusion-weighted imaging of the pancreas

BACKGROUND

It has been reported that deep learning-based reconstruction (DLR) can reduce image noise and artifacts, thereby improving the signal-to-noise ratio and image sharpness. However, no previous studies have evaluated the efficacy of DLR in improving image quality in reduced-field-of-view (reduced-FOV) diffusion-weighted imaging (DWI) [field-of-view optimized and constrained undistorted single-shot (FOCUS)] of the pancreas. We hypothesized that a combination of these techniques would improve DWI image quality without prolonging the scan time but would influence the apparent diffusion coefficient calculation.

AIM

To evaluate the efficacy of DLR for image quality improvement of FOCUS of the pancreas.

METHODS

This was a retrospective study evaluated 37 patients with pancreatic cystic lesions who underwent magnetic resonance imaging between August 2021 and October 2021. We evaluated three types of FOCUS examinations: FOCUS with DLR (FOCUS-DLR+), FOCUS without DLR (FOCUS-DLR-), and conventional FOCUS (FOCUS-conv). The three types of FOCUS and their apparent diffusion coefficient (ADC) maps were compared qualitatively and quantitatively.

RESULTS

FOCUS-DLR+ (3.62, average score of two radiologists) showed significantly better qualitative scores for image noise than FOCUS-DLR- (2.62) and FOCUS-conv (2.88) (P < 0.05). Furthermore, FOCUS-DLR+ showed the highest contrast ratio (CR) between the pancreatic parenchyma and adjacent fat tissue for b-values of 0 and 600 s/mm2 (0.72 ± 0.08 and 0.68 ± 0.08) and FOCUS-DLR- showed the highest CR between cystic lesions and the pancreatic parenchyma for the b-values of 0 and 600 s/mm2 (0.62 ± 0.21 and 0.62 ± 0.21) (P < 0.05), respectively. FOCUS-DLR+ provided significantly higher ADCs of the pancreas and lesion (1.44 ± 0.24 and 3.00 ± 0.66) compared to FOCUS-DLR- (1.39 ± 0.22 and 2.86 ± 0.61) and significantly lower ADCs compared to FOCUS-conv (1.84 ± 0.45 and 3.32 ± 0.70) (P < 0.05), respectively.

CONCLUSION

This study evaluated the efficacy of DLR for image quality improvement in reduced-FOV DWI of the pancreas. DLR can significantly denoise images without prolonging the scan time or decreasing the spatial resolution. The denoising level of DWI can be controlled to make the images appear more natural to the human eye. However, this study revealed that DLR did not ameliorate pancreatic distortion. Additionally, physicians should pay attention to the interpretation of ADCs after DLR application because ADCs are significantly changed by DLR.

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.

Usefulness of T2-Weighted Images with Deep-Learning-Based Reconstruction in Nasal Cartilage

OBJECTIVE

This study aims to evaluate the feasibility of visualizing nasal cartilage using deep-learning-based reconstruction (DLR) fast spin-echo (FSE) imaging in comparison to three-dimensional fast spoiled gradient-echo (3D FSPGR) images.

MATERIALS AND METHODS

This retrospective study included 190 set images of 38 participants, including axial T1- and T2-weighted FSE images using DLR (T1WIDL and T2WIDL, belong to FSEDL) and without using DLR (T1WIO and T2WIO, belong to FSEO) and 3D FSPGR images. Subjective evaluation (overall image quality, noise, contrast, artifacts, and identification of anatomical structures) was independently conducted by two radiologists. Objective evaluation including signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) was conducted using manual region-of-interest (ROI)-based analysis. Coefficient of variation (CV) and Bland-Altman plots were used to demonstrate the intra-rater repeatability of measurements for cartilage thickness on five different images.

RESULTS

Both qualitative and quantitative results confirmed superior FSEDL to 3D FSPGR images (both p < 0.05), improving the diagnosis confidence of the observers. Lower lateral cartilage (LLC), upper lateral cartilage (ULC), and septal cartilage (SP) were relatively well delineated on the T2WIDL, while 3D FSPGR showed poorly on the septal cartilage. For the repeatability of cartilage thickness measurements, T2WIDL showed the highest intra-observer (%CV = 8.7% for SP, 9.5% for ULC, and 9.7% for LLC) agreements. In addition, the acquisition time for T1WIDL and T2WIDL was respectively reduced by 14.2% to 29% compared to 3D FSPGR (both p < 0.05).

CONCLUSIONS

Two-dimensional equivalent-thin-slice T1- and T2-weighted images using DLR showed better image quality and shorter scan time than 3D FSPGR and conventional construction images in nasal cartilages. The anatomical details were preserved without losing clinical performance on diagnosis and prognosis, especially for pre-rhinoplasty planning.

Diffusion MRI data analysis assisted by deep learning synthesized anatomical images (DeepAnat)

Diffusion MRI is a useful neuroimaging tool for non-invasive mapping of human brain microstructure and structural connections. The analysis of diffusion MRI data often requires brain segmentation, including volumetric segmentation and cerebral cortical surfaces, from additional high-resolution T1-weighted (T1w) anatomical MRI data, which may be unacquired, corrupted by subject motion or hardware failure, or cannot be accurately co-registered to the diffusion data that are not corrected for susceptibility-induced geometric distortion. To address these challenges, this study proposes to synthesize high-quality T1w anatomical images directly from diffusion data using convolutional neural networks (CNNs) (entitled "DeepAnat"), including a U-Net and a hybrid generative adversarial network (GAN), and perform brain segmentation on synthesized T1w images or assist the co-registration using synthesized T1w images. The quantitative and systematic evaluations using data of 60 young subjects provided by the Human Connectome Project (HCP) show that the synthesized T1w images and results for brain segmentation and comprehensive diffusion analysis tasks are highly similar to those from native T1w data. The brain segmentation accuracy is slightly higher for the U-Net than the GAN. The efficacy of DeepAnat is further validated on a larger dataset of 300 more elderly subjects provided by the UK Biobank. Moreover, the U-Nets trained and validated on the HCP and UK Biobank data are shown to be highly generalizable to the diffusion data from Massachusetts General Hospital Connectome Diffusion Microstructure Dataset (MGH CDMD) acquired with different hardware systems and imaging protocols and therefore can be used directly without retraining or with fine-tuning for further improved performance. Finally, it is quantitatively demonstrated that the alignment between native T1w images and diffusion images uncorrected for geometric distortion assisted by synthesized T1w images substantially improves upon that by directly co-registering the diffusion and T1w images using the data of 20 subjects from MGH CDMD. In summary, our study demonstrates the benefits and practical feasibility of DeepAnat for assisting various diffusion MRI data analyses and supports its use in neuroscientific applications.

Perceptually Motivated Generative Model for Magnetic Resonance Image Denoising

Image denoising is an important preprocessing step in low-level vision problems involving biomedical images. Noise removal techniques can greatly benefit raw corrupted magnetic resonance images (MRI). It has been discovered that the MR data is corrupted by a mixture of Gaussian-impulse noise caused by detector flaws and transmission errors. This paper proposes a deep generative model (GenMRIDenoiser) for dealing with this mixed noise scenario. This work makes four contributions. To begin, Wasserstein generative adversarial network (WGAN) is used in model training to mitigate the problem of vanishing gradient, mode collapse, and convergence issues encountered while training a vanilla GAN. Second, a perceptually motivated loss function is used to guide the training process in order to preserve the low-level details in the form of high-frequency components in the image. Third, batch renormalization is used between the convolutional and activation layers to prevent performance degradation under the assumption of non-independent and identically distributed (non-iid) data. Fourth, global feature attention module (GFAM) is appended at the beginning and end of the parallel ensemble blocks to capture the long-range dependencies that are often lost due to the small receptive field of convolutional filters. The experimental results over synthetic data and MRI stack obtained from real MR scanners indicate the potential utility of the proposed technique across a wide range of degradation scenarios.

Convergence of Artificial Intelligence and Neuroscience towards the Diagnosis of Neurological Disorders-A Scoping Review

Artificial intelligence (AI) is a field of computer science that deals with the simulation of human intelligence using machines so that such machines gain problem-solving and decision-making capabilities similar to that of the human brain. Neuroscience is the scientific study of the struczture and cognitive functions of the brain. Neuroscience and AI are mutually interrelated. These two fields help each other in their advancements. The theory of neuroscience has brought many distinct improvisations into the AI field. The biological neural network has led to the realization of complex deep neural network architectures that are used to develop versatile applications, such as text processing, speech recognition, object detection, etc. Additionally, neuroscience helps to validate the existing AI-based models. Reinforcement learning in humans and animals has inspired computer scientists to develop algorithms for reinforcement learning in artificial systems, which enables those systems to learn complex strategies without explicit instruction. Such learning helps in building complex applications, like robot-based surgery, autonomous vehicles, gaming applications, etc. In turn, with its ability to intelligently analyze complex data and extract hidden patterns, AI fits as a perfect choice for analyzing neuroscience data that are very complex. Large-scale AI-based simulations help neuroscientists test their hypotheses. Through an interface with the brain, an AI-based system can extract the brain signals and commands that are generated according to the signals. These commands are fed into devices, such as a robotic arm, which helps in the movement of paralyzed muscles or other human parts. AI has several use cases in analyzing neuroimaging data and reducing the workload of radiologists. The study of neuroscience helps in the early detection and diagnosis of neurological disorders. In the same way, AI can effectively be applied to the prediction and detection of neurological disorders. Thus, in this paper, a scoping review has been carried out on the mutual relationship between AI and neuroscience, emphasizing the convergence between AI and neuroscience in order to detect and predict various neurological disorders.

Combination Use of Compressed Sensing and Deep Learning for Shoulder Magnetic Resonance Imaging With Various Sequences

OBJECTIVE

For compressed sensing (CS) to become widely used in routine magnetic resonance imaging (MRI), it is essential to improve image quality. This study aimed to evaluate the usefulness of combining CS and deep learning-based reconstruction (DLR) for various sequences in shoulder MRI.

METHODS

This retrospective study included 37 consecutive patients who underwent undersampled shoulder MRIs, including T1-weighted (T1WI), T2-weighted (T2WI), and fat-saturation T2-weighted (FS-T2WI) images. Images were reconstructed using the conventional wavelet-based denoising method (wavelet method) and a combination of wavelet and DLR-based denoising methods (hybrid-DLR method) for each sequence. The signal-to-noise ratio and contrast-to-noise ratio of the bone, muscle, and fat and the full width at half maximum of the shoulder joint were compared between the 2 image types. In addition, 2 board-certified radiologists scored the image noise, contrast, sharpness, artifacts, and overall image quality of the 2 image types on a 4-point scale.

RESULTS

The signal-to-noise ratios and contrast-to-noise ratios of the bone, muscle, and fat in T1WI, T2WI, and FS-T2WI obtained from the hybrid-DLR method were significantly higher than those of the conventional wavelet method (P < 0.001). However, there were no significant differences in the full width at half maximum of the shoulder joint in any of the sequences (P > 0.05). Furthermore, in all sequences, the mean scores of the image noise, sharpness, artifacts, and overall image quality were significantly higher in the hybrid-DLR method than in the wavelet method (P < 0.001), but there were no significant differences in contrast among the sequences (P > 0.05).

CONCLUSIONS

The DLR denoising method can improve the image quality of CS in T1-weighted images, T2-weighted images, and fat-saturation T2-weighted images of the shoulder compared with the wavelet denoising method alone.

Deep compressed sensing MRI via a gradient-enhanced fusion model

BACKGROUND

Compressed sensing has been employed to accelerate magnetic resonance imaging by sampling fewer measurements. However, conventional iterative optimization-based CS-MRI are time-consuming for iterative calculations and often share poor generalization ability on multicontrast datasets. Most deep-learning-based CS-MRI focus on learning an end-to-end mapping while ignoring some prior knowledge existed in MR images.

PURPOSE

We propose an iterative fusion model to integrate the image and gradient-based priors into reconstruction via convolutional neural network models while maintaining high quality and preserving the detailed information as well.

METHODS

We propose a gradient-enhanced fusion network (GFN) for fast and accurate MRI reconstruction, in which dense blocks with dilated convolution and dense residual learnings are used to capture abundant features with fewer parameters. Meanwhile, decomposed gradient maps containing obvious structural information are introduced into the network to enhance the reconstructed images. Besides this, gradient-based priors along directions X and Y are exploited to learn adaptive tight frames for reconstructing the desired image contrast and edges by respective gradient fusion networks. After that, both image and gradient priors are fused in the proposed optimization model, in which we employ the l₂ -norm to promote the sparsity of gradient priors. The proposed fusion model can effectively help to capture edge structures in the gradient images and to preserve more detailed information of MR images.

RESULTS

Experimental results demonstrate that the proposed method outperforms several CS-MRI methods in terms of peak signal-to-noise (PSNR), the structural similarity index (SSIM), and visualizations on three sampling masks with different rates. Noteworthy, to evaluate the generalization ability, the proposed model conducts cross-center training and testing experiments for all three modalities and shares more stable performance compared than other approaches. In addition, the proposed fusion model is applied to other comparable deep learning methods. The quantitative results show that the reconstruction results of these methods are obviously improved.

CONCLUSIONS

The gradient-based priors reconstructed from GFNs can effectively enhance the edges and details of under-sampled data. The proposed fusion model integrates image and gradient priors using l₂ -norm can effectively improve the generalization ability on multicontrast datasets reconstruction.

Deep Learning for Image Enhancement and Correction in Magnetic Resonance Imaging-State-of-the-Art and Challenges

Magnetic resonance imaging (MRI) provides excellent soft-tissue contrast for clinical diagnoses and research which underpin many recent breakthroughs in medicine and biology. The post-processing of reconstructed MR images is often automated for incorporation into MRI scanners by the manufacturers and increasingly plays a critical role in the final image quality for clinical reporting and interpretation. For image enhancement and correction, the post-processing steps include noise reduction, image artefact correction, and image resolution improvements. With the recent success of deep learning in many research fields, there is great potential to apply deep learning for MR image enhancement, and recent publications have demonstrated promising results. Motivated by the rapidly growing literature in this area, in this review paper, we provide a comprehensive overview of deep learning-based methods for post-processing MR images to enhance image quality and correct image artefacts. We aim to provide researchers in MRI or other research fields, including computer vision and image processing, a literature survey of deep learning approaches for MR image enhancement. We discuss the current limitations of the application of artificial intelligence in MRI and highlight possible directions for future developments. In the era of deep learning, we highlight the importance of a critical appraisal of the explanatory information provided and the generalizability of deep learning algorithms in medical imaging.

A two-stage framework for optical coherence tomography angiography image quality improvement

Introduction

Optical Coherence Tomography Angiography (OCTA) is a new non-invasive imaging modality that gains increasing popularity for the observation of the microvasculatures in the retina and the conjunctiva, assisting clinical diagnosis and treatment planning. However, poor imaging quality, such as stripe artifacts and low contrast, is common in the acquired OCTA and in particular Anterior Segment OCTA (AS-OCTA) due to eye microtremor and poor illumination conditions. These issues lead to incomplete vasculature maps that in turn makes it hard to make accurate interpretation and subsequent diagnosis.

Methods

In this work, we propose a two-stage framework that comprises a de-striping stage and a re-enhancing stage, with aims to remove stripe noise and to enhance blood vessel structure from the background. We introduce a new de-striping objective function in a Stripe Removal Net (SR-Net) to suppress the stripe noise in the original image. The vasculatures in acquired AS-OCTA images usually exhibit poor contrast, so we use a Perceptual Structure Generative Adversarial Network (PS-GAN) to enhance the de-striped AS-OCTA image in the re-enhancing stage, which combined cyclic perceptual loss with structure loss to achieve further image quality improvement.

Results and discussion

To evaluate the effectiveness of the proposed method, we apply the proposed framework to two synthetic OCTA datasets and a real AS-OCTA dataset. Our results show that the proposed framework yields a promising enhancement performance, which enables both conventional and deep learning-based vessel segmentation methods to produce improved results after enhancement of both retina and AS-OCTA modalities.

Application of artificial intelligence to imaging interpretations in the musculoskeletal area: Where are we? Where are we going?

The interest of researchers, clinicians and radiologists, in artificial intelligence (AI) continues to grow. Deep learning is a subset of machine learning, in which the computer algorithm itself can determine the optimal imaging features to answer a clinical question. Convolutional neural networks are the most common architecture for performing deep learning on medical images. The various musculoskeletal applications of deep learning are the detection of abnormalities on X-rays or cross-sectional images (CT, MRI), for example the detection of fractures, meniscal tears, anterior cruciate ligament tears, degenerative lesions of the spine, bone metastases, classification of e.g., dural sac stenosis, degeneration of intervertebral discs, assessment of skeletal age, and segmentation, for example of cartilage. Software developments are already impacting the daily practice of orthopedic imaging by automatically detecting fractures on radiographs. Improving image acquisition protocols, improving the quality of low-dose CT images, reducing acquisition times in MRI, or improving MR image resolution is possible through deep learning. Deep learning offers an automated way to offload time-consuming manual processes and improve practitioner performance. This article reviews the current state of AI in musculoskeletal imaging.

Denoise ultra-low-field 3D magnetic resonance images using a joint signal-image domain filter

Ultra-low-field magnetic resonance imaging (MRI) could suffer from heavy uncorrelated noise, and its removal could be a critical post-processing task. As a primary source of interference, Gaussian noise could corrupt the sampled MR signal (k-space data), especially at lower B₀ field strength. For this reason, we consider both signal and image domains by proposing a new joint filter characterized by a Kalman filter with linear prediction and a nonlocal mean filter with higher-order singular value decomposition (HOSVD) for denoising 3D MR data. The Kalman filter first attenuates the noise in k-space, and then its reconstruction images are used to guide HOSVD denoising process with exploring self-similarity among 3D structures. The clearer prefiltered images could also generate improved HOSVD learned bases used to transform the noise corrupted patch groups in the original MR data. The flexibility of proposed method is also demonstrated by integrating other k-space filters into the algorithm scheme. Experimental data includes simulated MR images with the varying noise level and real MR images obtained from our 50 mT MRI scanner. The results reveal that our method has a better noise-removal ability and introduces lesser unexpected artifacts than other related MRI denoising approaches.

Enhancement of 18F-Fluorodeoxyglucose PET Image Quality by Deep-Learning-Based Image Reconstruction Using Advanced Intelligent Clear-IQ Engine in Semiconductor-Based PET/CT Scanners

Deep learning (DL) image quality improvement has been studied for application to ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT). It is unclear, however, whether DL can increase the quality of images obtained with semiconductor-based PET/CT scanners. This study aimed to compare the quality of semiconductor-based PET/CT scanner images obtained by DL-based technology and conventional OSEM image with Gaussian postfilter. For DL-based data processing implementation, we used Advanced Intelligent Clear-IQ Engine (AiCE, Canon Medical Systems, Tochigi, Japan) and for OSEM images, Gaussian postfilter of 3 mm FWHM is used. Thirty patients who underwent semiconductor-based PET/CT scanner imaging between May 6, 2021, and May 19, 2021, were enrolled. We compared AiCE images and OSEM images and scored them for delineation, image noise, and overall image quality. We also measured standardized uptake values (SUVs) in tumors and healthy tissues and compared them between AiCE and OSEM. AiCE images scored significantly higher than OSEM images for delineation, image noise, and overall image quality. The Fleiss kappa value for the interobserver agreement was 0.57. Among the 21 SUV measurements in healthy organs, 11 (52.4%) measurements were significantly different between AiCE and OSEM images. More pathological lesions were detected in AiCE images as compared with OSEM images, with AiCE images showing higher SUVs for pathological lesions than OSEM images. AiCE can improve the quality of images acquired with semiconductor-based PET/CT scanners, including the noise level, contrast, and tumor detection capability.

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.

Denoising Using Noise2Void for Low-Field Magnetic Resonance Imaging: A Phantom Study

To reduce noise for low-field magnetic resonance imaging (MRI) using Noise2Void (N2V) and to demonstrate the N2V validity. N2V is one of the denoising convolutional neural network methods that allows the training of a model without a noiseless clean image. In this study, a kiwi fruit was scanned using a 0.35 Tesla MRI system, and the image qualities at pre- and postdenoising were evaluated. Structural similarity (SSIM), signal-to-noise ratio (SNR), and contrast ratio (CR) were measured, and visual assessment of noise and sharpness was observed. Both SSIM and SNR were significantly improved using N2V (P < 0.05). CR was unchanged between pre- and postdenoising images. The results of visual assessment for noise revealed higher scores in postdenoising images than that in predenoising images. The sharpness scores of postdenoising images were high when SNR was low. N2V provides effective noise reduction and is a useful denoising technique in low-field MRI.

Denoising diffusion weighted imaging data using convolutional neural networks

Diffusion weighted imaging (DWI) with multiple, high b-values is critical for extracting tissue microstructure measurements; however, high b-value DWI images contain high noise levels that can overwhelm the signal of interest and bias microstructural measurements. Here, we propose a simple denoising method that can be applied to any dataset, provided a low-noise, single-subject dataset is acquired using the same DWI sequence. The denoising method uses a one-dimensional convolutional neural network (1D-CNN) and deep learning to learn from a low-noise dataset, voxel-by-voxel. The trained model can then be applied to high-noise datasets from other subjects. We validated the 1D-CNN denoising method by first demonstrating that 1D-CNN denoising resulted in DWI images that were more similar to the noise-free ground truth than comparable denoising methods, e.g., MP-PCA, using simulated DWI data. Using the same DWI acquisition but reconstructed with two common reconstruction methods, i.e. SENSE1 and sum-of-square, to generate a pair of low-noise and high-noise datasets, we then demonstrated that 1D-CNN denoising of high-noise DWI data collected from human subjects showed promising results in three domains: DWI images, diffusion metrics, and tractography. In particular, the denoised images were very similar to a low-noise reference image of that subject, more than the similarity between repeated low-noise images (i.e. computational reproducibility). Finally, we demonstrated the use of the 1D-CNN method in two practical examples to reduce noise from parallel imaging and simultaneous multi-slice acquisition. We conclude that the 1D-CNN denoising method is a simple, effective denoising method for DWI images that overcomes some of the limitations of current state-of-the-art denoising methods, such as the need for a large number of training subjects and the need to account for the rectified noise floor.

Performance analysis of reshaped Gabor filter for removing the Rician distributed noise in brain MR images

Magnetic Resonance Imaging (MRI) is an essential clinical tool for detecting the abnormalities such as tumors and clots in the human brain. The brain MR images are contaminated by artifacts and noise that follow Rician distribution during the acquisition process. It causes the loss of fine details information, distortion, and a blurred vision of the image. A reshaped Gabor filter-based denoising technique is proposed to overcome these issues. To develop the reshaped Gabor filter, the range of reshaping parameters of the filter is initially obtained by a random search method. Further, to evaluate the better performance of the proposed filter, a manual search is used to find the optimal parametric values and tested on T1, T2, and PD weighted MR data sets one by one. Also, the proposed technique is compared with the existing state of the art filtering methods such as Wiener, Median, Partial differential equation (PDE), Anisotropic diffusion filter (ADF), Non-local means filter (NLM), Modified complex diffusion filter (MCD), Multichannel residual learning of CNN (MRL), Maximum a posteriori (MAP), Adaptive non-local means algorithm (ADNLM), and Advance NLM filtering with non-sub sampled (AVNLMNS) on the basic reference and no reference parameter. The parameters such as mean square error (MSE), peak signal to noise ratio (PSNR), structural similarity index metric (SSIM), perception-based image quality evaluator (PIQE), and blind/referenceless image spatial quality evaluator (BRISQE) are evaluated on T1, T2, and PD weighted MR images with different noise variances such as 1%, 3%, 5%, 7%, and 9%. The proposed method may be used as a better denoising scheme for Rician distributed noise, edge preservation, fine details restoration, and enhancement of abnormalities.

Research on Optimization Scheme for Blocking Artifacts after Patch-Based Medical Image Reconstruction

Due to limitations of computer resources, when utilizing a neural network to process an image with a high resolution, the typical processing approach is to slice the original image. However, because of the influence of zero-padding in the edge component during the convolution process, the central part of the patch often has more accurate feature information than the edge part, resulting in image blocking artifacts after patch stitching. We studied this problem in this paper and proposed a fusion method that assigns a weight to each pixel in a patch using a truncated Gaussian function as the weighting function. In this method, we used the weighting function to transform the Euclidean-distance between a point in the overlapping part and the central point of the patch where the point was located into a weight coefficient. With increasing distance, the value of the weight coefficient decreased. Finally, the reconstructed image was obtained by weighting. We employed the bias correction model to evaluate our method on the simulated database BrainWeb and the real dataset HCP (Human Connectome Project). The results show that the proposed method is capable of effectively removing blocking artifacts and obtaining a smoother bias field. To verify the effectiveness of our algorithm, we employed a denoising model to test it on the IXI-Guys human dataset. Qualitative and quantitative evaluations of both models show that the fusion method proposed in this paper can effectively remove blocking artifacts and demonstrates superior performance compared to five commonly available and state-of-the-art fusion methods.

SM-SegNet: A Lightweight Squeeze M-SegNet for Tissue Segmentation in Brain MRI Scans

In this paper, we propose a novel squeeze M-SegNet (SM-SegNet) architecture featuring a fire module to perform accurate as well as fast segmentation of the brain on magnetic resonance imaging (MRI) scans. The proposed model utilizes uniform input patches, combined-connections, long skip connections, and squeeze-expand convolutional layers from the fire module to segment brain MRI data. The proposed SM-SegNet architecture involves a multi-scale deep network on the encoder side and deep supervision on the decoder side, which uses combined-connections (skip connections and pooling indices) from the encoder to the decoder layer. The multi-scale side input layers support the deep network layers' extraction of discriminative feature information, and the decoder side provides deep supervision to reduce the gradient problem. By using combined-connections, extracted features can be transferred from the encoder to the decoder resulting in recovering spatial information, which makes the model converge faster. Long skip connections were used to stabilize the gradient updates in the network. Owing to the adoption of the fire module, the proposed model was significantly faster to train and offered a more efficient memory usage with 83% fewer parameters than previously developed methods, owing to the adoption of the fire module. The proposed method was evaluated using the open-access series of imaging studies (OASIS) and the internet brain segmentation registry (IBSR) datasets. The experimental results demonstrate that the proposed SM-SegNet architecture achieves segmentation accuracies of 95% for cerebrospinal fluid, 95% for gray matter, and 96% for white matter, which outperforms the existing methods in both subjective and objective metrics in brain MRI segmentation.

Multi-modal trained artificial intelligence solution to triage chest X-ray for COVID-19 using pristine ground-truth, versus radiologists

The front-line imaging modalities computed tomography (CT) and X-ray play important roles for triaging COVID patients. Thoracic CT has been accepted to have higher sensitivity than a chest X-ray for COVID diagnosis. Considering the limited access to resources (both hardware and trained personnel) and issues related to decontamination, CT may not be ideal for triaging suspected subjects. Artificial intelligence (AI) assisted X-ray based application for triaging and monitoring require experienced radiologists to identify COVID patients in a timely manner with the additional ability to delineate and quantify the disease region is seen as a promising solution for widespread clinical use. Our proposed solution differs from existing solutions presented by industry and academic communities. We demonstrate a functional AI model to triage by classifying and segmenting a single chest X-ray image, while the AI model is trained using both X-ray and CT data. We report on how such a multi-modal training process improves the solution compared to single modality (X-ray only) training. The multi-modal solution increases the AUC (area under the receiver operating characteristic curve) from 0.89 to 0.93 for a binary classification between COVID-19 and non-COVID-19 cases. It also positively impacts the Dice coefficient (0.59 to 0.62) for localizing the COVID-19 pathology. To compare the performance of experienced readers to the AI model, a reader study is also conducted. The AI model showed good consistency with respect to radiologists. The DICE score between two radiologists on the COVID group was 0.53 while the AI had a DICE value of 0.52 and 0.55 when compared to the segmentation done by the two radiologists separately. From a classification perspective, the AUCs of two readers was 0.87 and 0.81 while the AUC of the AI is 0.93 based on the reader study dataset. We also conducted a generalization study by comparing our method to the-state-art methods on independent datasets. The results show better performance from the proposed method. Leveraging multi-modal information for the development benefits the single-modal inferencing.

From Dose Reduction to Contrast Maximization: Can Deep Learning Amplify the Impact of Contrast Media on Brain Magnetic Resonance Image Quality? A Reader Study

OBJECTIVES

The aim of this study was to evaluate a deep learning method designed to increase the contrast-to-noise ratio in contrast-enhanced gradient echo T1-weighted brain magnetic resonance imaging (MRI) acquisitions. The processed images are quantitatively evaluated in terms of lesion detection performance.

MATERIALS AND METHODS

A total of 250 multiparametric brain MRIs, acquired between November 2019 and March 2021 at Gustave Roussy Cancer Campus (Villejuif, France), were considered for inclusion in this retrospective monocentric study. Independent training (107 cases; age, 55 ± 14 years; 58 women) and test (79 cases; age, 59 ± 14 years; 41 women) samples were defined. Patients had glioma, brain metastasis, meningioma, or no enhancing lesion. Gradient echo and turbo spin echo with variable flip angles postcontrast T1 sequences were acquired in all cases. For the cases that formed the training sample, "low-dose" postcontrast gradient echo T1 images using 0.025 mmol/kg injections of contrast agent were also acquired. A deep neural network was trained to synthetically enhance the low-dose T1 acquisitions, taking standard-dose T1 MRI as reference. Once trained, the contrast enhancement network was used to process the test gradient echo T1 images. A read was then performed by 2 experienced neuroradiologists to evaluate the original and processed T1 MRI sequences in terms of contrast enhancement and lesion detection performance, taking the turbo spin echo sequences as reference.

RESULTS

The processed images were superior to the original gradient echo and reference turbo spin echo T1 sequences in terms of contrast-to-noise ratio (44.5 vs 9.1 and 16.8; P < 0.001), lesion-to-brain ratio (1.66 vs 1.31 and 1.44; P < 0.001), and contrast enhancement percentage (112.4% vs 85.6% and 92.2%; P < 0.001) for cases with enhancing lesions. The overall image quality of processed T1 was preferred by both readers (graded 3.4/4 on average vs 2.7/4; P < 0.001). Finally, the proposed processing improved the average sensitivity of gradient echo T1 MRI from 88% to 96% for lesions larger than 10 mm (P = 0.008), whereas no difference was found in terms of the false detection rate (0.02 per case in both cases; P > 0.99). The same effect was observed when considering all lesions larger than 5 mm: sensitivity increased from 70% to 85% (P < 0.001), whereas false detection rates remained similar (0.04 vs 0.06 per case; P = 0.48). With all lesions included regardless of their size, sensitivities were 59% and 75% for original and processed T1 images, respectively (P < 0.001), and the corresponding false detection rates were 0.05 and 0.14 per case, respectively (P = 0.06).

CONCLUSION

The proposed deep learning method successfully amplified the beneficial effects of contrast agent injection on gradient echo T1 image quality, contrast level, and lesion detection performance. In particular, the sensitivity of the MRI sequence was improved by up to 16%, whereas the false detection rate remained similar.

Deep Learning-Based Ultrasound Combined with Gastroscope for the Diagnosis and Nursing of Upper Gastrointestinal Submucous Lesions

The study focused on the diagnostic value of deep learning-based ultrasound combined with gastroscope examination for upper gastrointestinal submucous lesions and nursing. A total of 104 patients with upper gastrointestinal submucous lesions diagnosed in hospital were selected as the research subjects. In this study, the feed forward denoising convulsive neural network (DnCNN) was improved, and the n-DnCNN model was designed and applied to ultrasonic image processing. The peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) of Gaussian filtering, NL-means, and DnCNN were then compared with n-DnCNN. Subsequently, the distribution and types of submucosal lesions in different parts of the upper digestive tract were analyzed by ultrasound combined with gastroscope and gastroscope examination alone, and the diagnostic performance of this method was evaluated. The results showed that the average PSNR and SSIM of the n-DnCNN model were 33.01 dB and 0.87, respectively, which were significantly higher than GF, NL-means, and DnCNN algorithms, and the difference was statistically significant ( $P<0.05$ ). Of the 116 lesions detected, 49 were located in the esophagus (42.24%), 52 in the stomach (44.83%), and 15 in the duodenum (12.93%). Of the 49 esophageal submucosal lesions, 6.12% were located in the upper esophagus, 55.1% in the middle esophagus, and 38.79% in the lower esophagus, and the difference was statistically significant ( $P<0.05$ ). Of the gastric submucosal lesions, the lesions in the gastric cardia were significantly less than in other parts, and the difference was statistically significant ( $P<0.05$ ). The accuracy of ultrasound combined with gastroscope in the diagnosis of upper gastrointestinal submucous episodes was 82.32%, higher than that of gastroscope examination, and the difference was statistically significant ( $P<0.05$ ). In conclusion, the n-DnCNN model has a good noise reduction effect, and the obtained image is of high quality. Ultrasound combined with gastroscope examination can effectively improve the accuracy of diagnosis of upper gastrointestinal submucous lesions.

Deep learning-based velocity antialiasing of 4D-flow MRI

Purpose

To develop a convolutional neural network (CNN) for the robust and fast correction of velocity aliasing in 4D‐flow MRI.

Methods

This study included 667 adult subjects with aortic 4D‐flow MRI data with existing velocity aliasing (n = 362) and no velocity aliasing (n = 305). Additionally, 10 controls received back‐to‐back 4D‐flow scans with systemically varied velocity‐encoding sensitivity (vencs) at 60, 100, and 175 cm/s. The no‐aliasing data sets were used to simulate velocity aliasing by reducing the venc to 40%–70% of the original, alongside a ground truth locating all aliased voxels (153 training, 152 testing). The 152 simulated and 362 existing aliasing data sets were used for testing and compared with a conventional velocity antialiasing algorithm. Dice scores were calculated to quantify CNN performance. For controls, the venc 175‐cm/s scans were used as the ground truth and compared with the CNN‐corrected venc 60 and 100 cm/s data sets

Results

The CNN required 176 ± 30 s to perform compared with 162 ± 14 s for the conventional algorithm. The CNN showed excellent performance for the simulated data compared with the conventional algorithm (median range of Dice scores CNN: [0.89–0.99], conventional algorithm: [0.84–0.94], p < 0.001, across all simulated vencs) and detected more aliased voxels in existing velocity aliasing data sets (median detected CNN: 159 voxels [31–605], conventional algorithm: 65 [7–417], p < 0.001). For controls, the CNN showed Dice scores of 0.98 [0.95–0.99] and 0.96 [0.87–0.99] for venc = 60 cm/s and 100 cm/s, respectively, while flow comparisons showed moderate‐excellent agreement.

Conclusion

Deep learning enabled fast and robust velocity anti‐aliasing in 4D‐flow MRI.

SDnDTI: Self-supervised deep learning-based denoising for diffusion tensor MRI

Diffusion tensor magnetic resonance imaging (DTI) is a widely adopted neuroimaging method for the in vivo mapping of brain tissue microstructure and white matter tracts. Nonetheless, the noise in the diffusion-weighted images (DWIs) decreases the accuracy and precision of DTI derived microstructural parameters and leads to prolonged acquisition time for achieving improved signal-to-noise ratio (SNR). Deep learning-based image denoising using convolutional neural networks (CNNs) has superior performance but often requires additional high-SNR data for supervising the training of CNNs, which reduces the feasibility of supervised learning-based denoising in practice. In this work, we develop a self-supervised deep learning-based method entitled "SDnDTI" for denoising DTI data, which does not require additional high-SNR data for training. Specifically, SDnDTI divides multi-directional DTI data into many subsets of six DWI volumes and transforms DWIs from each subset to along the same diffusion-encoding directions through the diffusion tensor model, generating multiple repetitions of DWIs with identical image contrasts but different noise observations. SDnDTI removes noise by first denoising each repetition of DWIs using a deep 3-dimensional CNN with the average of all repetitions with higher SNR as the training target, following the same approach as normal supervised learning based denoising methods, and then averaging CNN-denoised images for achieving higher SNR. The denoising efficacy of SDnDTI is demonstrated in terms of the similarity of output images and resultant DTI metrics compared to the ground truth generated using substantially more DWI volumes on two datasets with different spatial resolutions, b-values and numbers of input DWI volumes provided by the Human Connectome Project (HCP) and the Lifespan HCP in Aging. The SDnDTI results preserve image sharpness and textural details and substantially improve upon those from the raw data. The results of SDnDTI are comparable to those from supervised learning-based denoising and outperform those from state-of-the-art conventional denoising algorithms including BM4D, AONLM and MPPCA. By leveraging domain knowledge of diffusion MRI physics, SDnDTI makes it easier to use CNN-based denoising methods in practice and has the potential to benefit a wider range of research and clinical applications that require accelerated DTI acquisition and high-quality DTI data for mapping of tissue microstructure, fiber tracts and structural connectivity in the living human brain.

Improved magnetic resonance myelin water imaging using multi-channel denoising convolutional neural networks (MCDnCNN)

BACKGROUND

Myelin water imaging (MWI) is powerful and important for studying and diagnosing neurological and psychiatric diseases. In particular, myelin water fraction (MWF) is derived from MWI data for quantifying myelination. However, MWF estimation is typically sensitive to noise. Improving the accuracy of MWF estimation based on WMI data acquired using a magnetic resonance (MR) multiple gradient recalled echo (mGRE) imaging sequence is desired.

METHODS

The proposed method employs a recently introduced the multi-channel denoising convolutional neural networks (MCDnCNN). Five different MCDnCNN models, denoted as Delevel1, Delevel2, Delevel3, Delevel4 and DelevelMix corresponding to five noise levels (Level1, Level2, Level3, Level4 and LevelMix), were trained using the data of the first echo of the mGRE brain images acquired from 15 healthy human subjects. Using simulated noisy data that employed a hollow cylinder model, we first evaluated the improvement in estimating MWF based on data denoised by the five different MCDnCNNs, by comparing the MWF maps calculated from the denoised data with ground truth. Next, we again evaluated the improvement using real-world in vivo datasets of 11 human participants acquired using the mGRE sequence. The datasets were first denoised by five different MCDnCNNs (Delevel1, 2, 3, 4 and DelevelMix), and subsequently their MWF maps were calculated and compared with the MWF maps directly calculated from the raw mGRE images without being denoised.

RESULTS

Experiments using the simulation data denoised by the appropriate MCDnCNN models showed that the standard deviation (SD) of the absolute error (AE) of the derived MWF results was significantly reduced (maximal reduction =15.5%, Level3 simulated noisy data, orientation angle =0, all the five MCDnCNN models). In the test using in vivo data, estimating MWF based on data particularly denoised by the appropriate MCDnCNN models was found to be the best, compared to otherwise not using the appropriate models. The results demonstrated that the appropriate MCDnCNN models may permit high-quality MWF mapping, i.e., substantial reduction of random variation in estimating MWF-maps while preserving accuracy and structural details.

CONCLUSIONS

Appropriate MCDnCNN models as proposed may improve both the accuracy and precision in estimating MWF maps, thereby making it a more clinically feasible alternative.

Deep regression with ensembles enables fast, first-order shimming in low-field NMR

Shimming in the context of nuclear magnetic resonance aims to achieve a uniform magnetic field distribution, as perfect as possible, and is crucial for useful spectroscopy and imaging. Currently, shimming precedes most acquisition procedures in the laboratory, and this mostly semi-automatic procedure often needs to be repeated, which can be cumbersome and time-consuming. The paper investigates the feasibility of completely automating and accelerating the shimming procedure by applying deep learning (DL). We show that DL can relate measured spectral shape to shim current specifications and thus rapidly predict three shim currents simultaneously, given only four input spectra. Due to the lack of accessible data for developing shimming algorithms, we also introduce a database that served as our DL training set, and allows inference of changes to ¹H NMR signals depending on shim offsets. In situ experiments of deep regression with ensembles demonstrate a high success rate in spectral quality improvement for random shim distortions over different neural architectures and chemical substances. This paper presents a proof-of-concept that machine learning can simplify and accelerate the shimming problem, either as a stand-alone method, or in combination with traditional shimming methods. Our database and code are publicly available.

Hybrid deep-learning-based denoising method for compressed sensing in pituitary MRI: comparison with the conventional wavelet-based denoising method

OBJECTIVES

This study aimed to evaluate the efficacy of a combined wavelet and deep-learning reconstruction (DLR) method for under-sampled pituitary MRI.

METHODS

This retrospective study included 28 consecutive patients who underwent under-sampled pituitary T2-weighted images (T2WI). Images were reconstructed using either the conventional wavelet denoising method (wavelet method) or the wavelet and DLR methods combined (hybrid DLR method) at five denoising levels. The signal-to-noise ratio (SNR) of the CSF, hypothalamic, and pituitary images and the contrast between structures were compared between the two image types. Noise quality, contrast, sharpness, artifacts, and overall image quality were evaluated by two board-certified radiologists. The quantitative and the qualitative analyses were performed with robust two-way repeated analyses of variance.

RESULTS

Using the hybrid DLR method, the SNR of the CSF progressively increased as denoising levels increased. By contrast, with the wavelet method, the SNR of the CSF, hypothalamus, and pituitary did not increase at higher denoising levels. There was a significant main effect of denoising methods (p < 0.001) and denoising levels (p < 0.001), and an interaction between denoising methods and denoising levels (p < 0.001). For all five qualitative scores, there was a significant main effect of denoising methods (p < 0.001) and an interaction between denoising methods and denoising levels (p < 0.001).

CONCLUSIONS

The hybrid DLR method can provide higher image quality for T2WI of the pituitary with compressed sensing (CS) than the wavelet method alone, especially at higher denoising levels.

KEY POINTS

• The signal-to-noise ratios of cerebrospinal fluid progressively increased with the hybrid DLR method, with an increase in the denoising level for cerebrospinal fluid in pituitary T2WI with CS. • The signal-to-noise ratios of cerebrospinal fluid using the conventional wavelet method did not increase at higher denoising levels. • All qualitative scores of hybrid deep-learning reconstructions at all denoising levels were higher than those for the wavelet denoising method.

Content-Noise Complementary Learning for Medical Image Denoising

Medical imaging denoising faces great challenges, yet is in great demand. With its distinctive characteristics, medical imaging denoising in the image domain requires innovative deep learning strategies. In this study, we propose a simple yet effective strategy, the content-noise complementary learning (CNCL) strategy, in which two deep learning predictors are used to learn the respective content and noise of the image dataset complementarily. A medical image denoising pipeline based on the CNCL strategy is presented, and is implemented as a generative adversarial network, where various representative networks (including U-Net, DnCNN, and SRDenseNet) are investigated as the predictors. The performance of these implemented models has been validated on medical imaging datasets including CT, MR, and PET. The results show that this strategy outperforms state-of-the-art denoising algorithms in terms of visual quality and quantitative metrics, and the strategy demonstrates a robust generalization capability. These findings validate that this simple yet effective strategy demonstrates promising potential for medical image denoising tasks, which could exert a clinical impact in the future. Code is available at: https://github.com/gengmufeng/CNCL-denoising.

AI MSK clinical applications: spine imaging

Recent investigations have focused on the clinical application of artificial intelligence (AI) for tasks specifically addressing the musculoskeletal imaging routine. Several AI applications have been dedicated to optimizing the radiology value chain in spine imaging, independent from modality or specific application. This review aims to summarize the status quo and future perspective regarding utilization of AI for spine imaging. First, the basics of AI concepts are clarified. Second, the different tasks and use cases for AI applications in spine imaging are discussed and illustrated by examples. Finally, the authors of this review present their personal perception of AI in daily imaging and discuss future chances and challenges that come along with AI-based solutions.

[Effects of Image-based Noise Reduction Software on Magnetic Resonance Imaging]

In magnetic resonance imaging (MRI) examinations, a trade-off is noted among acquisition time, resolution, and signal-to-noise ratio (SNR). High-resolution images are expected to improve the detection of small lesions. However, to ensure a high SNR, the imaging time must be extended. If the number of additions is reduced to shorten the imaging time, a reduction in slice thickness and in-plane resolution is necessary to ensure an adequate SNR. A combination of acceleration and denoising using deep learning has been previously reported. However, although it may be useful as a noise reduction technique onboard device, it cannot be used for general purposes. We studied the effects of a recently developed general-purpose image-based noise reduction software on MRI by measuring SNR and other parameters such as contrast, resolution, and noise power spectrum (NPS). NPS was influenced by the difference in processing mode, whereas contrast remained uninfluenced. Regarding resolution, the edge information was retained and was found to be better in iNoir 3D than in iNoir 2D. However, owing to the increased intensity of noise-reduction processing, the slope of the edge in the low-contrast area was smoothed, presenting a visually blurred impression.

Automatic breast tissue segmentation in MRIs with morphology snake and deep denoiser training via extended Stein's unbiased risk estimator

Accurate segmentation of the breast tissue is a significant challenge in the analysis of breast MR images, especially analysis of breast images with low contrast. Most of the existing methods for breast segmentation are semi-automatic and limited in their ability to achieve accurate results. This is because of difficulties in removing landmarks from noisy magnetic resonance images (MRI). Especially, when tumour is imaged for scanning, how to isolate the tumour region from chest will directly affect the accuracy for tumour to be detected. Due to low intensity levels and the close connection between breast and chest portion in MRIs, this study proposes an innovative, fully automatic and fast segmentation approach which combines histogram with inverse Gaussian gradient for morphology snakes, along with extended Stein's unbiased risk estimator (eSURE) applied for unsupervised learning of deep neural network Gaussian denoisers, aimed at accurate identification of landmarks such as chest and breast.

Introducing Swish and Parallelized Blind Removal Improves the Performance of a Convolutional Neural Network in Denoising MR Images

Purpose

To improve the performance of a denoising convolutional neural network (DnCNN) and to make it applicable to images with inhomogeneous noise, a refinement involving an activation function (AF) and an application of the refined method for inhomogeneous-noise images was examined in combination with parallelized image denoising.

Methods

Improvements in the DnCNN were performed by three approaches. One is refinement of the AF of each neural network that constructs the DnCNN. Swish was used in the DnCNN instead of rectifier linear unit. Second, blind noise removal was introduced to the DnCNN in order to adapt spatially variant noises. Third, blind noise removal was applied to parallelized image denoising, referred to herein as ParBID. The ParBID procedure is as follows: (1) adjacent 2D slice images are linearly combined to obtained higher peak SNR (PSNR) images, (2) combined images with different weight coefficients are denoised using the blind DnCNN, and (3) denoised combined images are separated into original position images by algebraic calculation.

Results

Experimental studies showed that the PSNR and the structural similarity index (SSIM) were improved by using Swish for all noise levels, from 2.5% to 7.5%, as compared to the conventional DnCNN. It was also shown that a well-trained CNN could remove spatially variant noises superimposed on images. Experimental studies with ParBID showed that the greatest PSNR and SSIM improvements were obtained at the middle slice when three slice images were used for linear image combination. More fine structures of images and image contrast remained when the proposed ParBID procedure was used.

Conclusion

Swish can improve the denoising performance of the DnCNN, and the denoising performance and effectiveness were further improved by ParBID.

Lesion Segmentation in Ultrasound Using Semi-Pixel-Wise Cycle Generative Adversarial Nets

Breast cancer is the most common invasive cancer with the highest cancer occurrence in females. Handheld ultrasound is one of the most efficient ways to identify and diagnose the breast cancer. The area and the shape information of a lesion is very helpful for clinicians to make diagnostic decisions. In this study we propose a new deep-learning scheme, semi-pixel-wise cycle generative adversarial net (SPCGAN) for segmenting the lesion in 2D ultrasound. The method takes the advantage of a fully convolutional neural network (FCN) and a generative adversarial net to segment a lesion by using prior knowledge. We compared the proposed method to a fully connected neural network and the level set segmentation method on a test dataset consisting of 32 malignant lesions and 109 benign lesions. Our proposed method achieved a Dice similarity coefficient (DSC) of 0.92 while FCN and the level set achieved 0.90 and 0.79 respectively. Particularly, for malignant lesions, our method increases the DSC (0.90) of the fully connected neural network to 0.93 significantly (p 0.001). The results show that our SPCGAN can obtain robust segmentation results. The framework of SPCGAN is particularly effective when sufficient training samples are not available compared to FCN. Our proposed method may be used to relieve the radiologists' burden for annotation.

Reconstruction of cardiovascular black-blood T2-weighted image by deep learning algorithm: A comparison with intensity filter

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.